Running Pv3Rs for Solomon Islands data
Shazia Ruybal-Pesántez
2024-08-27
Set global variables
From TO/JS - This section sets some flags that can be tweaked to re-run the analysis with different filtering parameters:
RUN_EXAMPLE: allows discarding alleles that have a
frequency below a threshold (defaults to FALSE),
WITHIN_INDIVIDUAL_ALLELE_FREQ_THR: allows discarding
alleles that have a frequency below a threshold (defaults to 0),
BENCHMARK_MARKERS: list of markers to include in the
analysis (defaults to the full list established by Jason, could be
reduced down to 3 for memory optimization),
MAX_MOI_TO_INCLUDE: allows discarding participant data when
they exhibit very complex infection patterns, counted as the total
number of clones across all infection events (defaults to 8),
PRIOR_3RS: a named vector with prior probabilities of
recrudescence (C), relapse (L) or reinfection (I) (defaults to uniform,
i.e. 1/3 each, but tweaked to minimize recrudescence in the present
study).
## RUN_EXAMPLE
## Boolean flag to enable/disable running the example Pv3Rs posterior
## computation on a single episode sampled at random from the dataset.
## DEFAULT: FALSE
RUN_EXAMPLE <- TRUE
## WITHIN_INDIVIDUAL_ALLELE_FREQ_THR
## Minimum threshold above which an allele is preserved in individual
## haplotype data to be preserved when reconstructing infection
## history. Note that this will *not* discard these alleles from the
## population-level allele frequency that is derived from the initial
## visit, but only discard that allele from individual observations
## when it is 'too rare to be exploited'.
## DEFAULT: 0
WITHIN_INDIVIDUAL_ALLELE_FREQ_THR <- 0
## BENCHMARK_MARKERS
## Vector of character string with names matching those from the
## markers of interest for amplicon sequencing. Only the markers
## listed in that vector will be preserved in individual-level data
## and identity of infection will be solely based on the alleles
## observed for these markers.
## The markers are listed by importance as discovered by Jason.
## DEFAULT: all markers (could be suboptimal/too memory-consuming?)
BENCHMARK_MARKERS <- c(
"Chr05",
"Chr07",
"Chr09",
"Chr10",
"Chr08",
"Chr13",
"Chr11",
"Chr03",
"Chr01",
"Chr02",
"Chr14"
)
## MAX_MOI_TO_INCLUDE
## As per Aimee's guidelines:
## We do not recommend running compute_posterior() for data
## whose total genotype count exceeds eight, where the total
## genotype count is the sum of per-episode maximum per-marker
## allele counts.
## The MAX_MOI_TO_INCLUDE expects an integer that will discard all
## individuals having a summed MOI > 8 across all recorded episodes.
## DEFAULT: 8
MAX_MOI_TO_INCLUDE <- 8
## PRIORS_3RS
## A vector of probabilities, summing to 1, corresponding to
## the probability of each stage for the 3Rs for Pv episodes.
## The vector order is re(C)rudescence, re(L)apse, re(I)nfection.
## In this clinical trial, we assume recrudescence is possible so,
## we use the default priors
## DEFAULT: c("C" = 1/3, "L" = 1/3, "I" = 1/3)
PRIOR_3RS <- c("C" = 1/3, "L" = 1/3, "I" = 1/3)
# Note: the below probabilities were used for SeroTAT study because it does not involve treatment at baseline
# PRIOR_3RS <- c("C" = 0.10, "L" = 0.45, "I" = 0.45)Solomon Islands data curation
Read in data from Solomon Islands:
sols <- read.csv(here("data/final", "sols_raw_data.csv"))
# head(sols)
# names(sols)Number of unique haplotypes (ie marker cardinality):
sols %>%
group_by(marker_id) %>%
summarise(unique_haplotypes = n_distinct(Haplotype))## # A tibble: 11 × 2
## marker_id unique_haplotypes
## <chr> <int>
## 1 Chr01 4
## 2 Chr02 10
## 3 Chr03 6
## 4 Chr05 13
## 5 Chr07 9
## 6 Chr08 2
## 7 Chr09 7
## 8 Chr10 6
## 9 Chr11 3
## 10 Chr13 5
## 11 Chr14 6Counts of each haplotype:
sols %>% count(marker_id, Haplotype)## marker_id Haplotype n
## 1 Chr01 Chr01-1 76
## 2 Chr01 Chr01-2 48
## 3 Chr01 Chr01-3 18
## 4 Chr01 Chr01-4 5
## 5 Chr02 Chr02-1 27
## 6 Chr02 Chr02-11 3
## 7 Chr02 Chr02-2 25
## 8 Chr02 Chr02-3 28
## 9 Chr02 Chr02-4 29
## 10 Chr02 Chr02-5 25
## 11 Chr02 Chr02-6 13
## 12 Chr02 Chr02-7 10
## 13 Chr02 Chr02-8 6
## 14 Chr02 Chr02-9 4
## 15 Chr03 Chr03-1 84
## 16 Chr03 Chr03-2 36
## 17 Chr03 Chr03-3 9
## 18 Chr03 Chr03-4 18
## 19 Chr03 Chr03-5 8
## 20 Chr03 Chr03-6 2
## 21 Chr05 Chr05-1 26
## 22 Chr05 Chr05-10 2
## 23 Chr05 Chr05-11 4
## 24 Chr05 Chr05-12 2
## 25 Chr05 Chr05-13 3
## 26 Chr05 Chr05-2 8
## 27 Chr05 Chr05-3 17
## 28 Chr05 Chr05-4 16
## 29 Chr05 Chr05-5 16
## 30 Chr05 Chr05-6 13
## 31 Chr05 Chr05-7 13
## 32 Chr05 Chr05-8 13
## 33 Chr05 Chr05-9 5
## 34 Chr07 Chr07-1 57
## 35 Chr07 Chr07-10 3
## 36 Chr07 Chr07-11 2
## 37 Chr07 Chr07-2 39
## 38 Chr07 Chr07-3 33
## 39 Chr07 Chr07-4 18
## 40 Chr07 Chr07-5 11
## 41 Chr07 Chr07-6 8
## 42 Chr07 Chr07-9 3
## 43 Chr08 Chr08-1 108
## 44 Chr08 Chr08-2 23
## 45 Chr09 Chr09-1 80
## 46 Chr09 Chr09-2 22
## 47 Chr09 Chr09-3 34
## 48 Chr09 Chr09-4 18
## 49 Chr09 Chr09-5 9
## 50 Chr09 Chr09-7 2
## 51 Chr09 Chr09-9 2
## 52 Chr10 Chr10-1 66
## 53 Chr10 Chr10-2 52
## 54 Chr10 Chr10-3 16
## 55 Chr10 Chr10-4 11
## 56 Chr10 Chr10-5 6
## 57 Chr10 Chr10-6 5
## 58 Chr11 Chr11-1 78
## 59 Chr11 Chr11-2 39
## 60 Chr11 Chr11-3 33
## 61 Chr13 Chr13-1 40
## 62 Chr13 Chr13-2 55
## 63 Chr13 Chr13-3 45
## 64 Chr13 Chr13-4 5
## 65 Chr13 Chr13-5 4
## 66 Chr14 Chr14-1 51
## 67 Chr14 Chr14-2 46
## 68 Chr14 Chr14-3 25
## 69 Chr14 Chr14-4 21
## 70 Chr14 Chr14-5 16
## 71 Chr14 Chr14-6 9“MOI”:
sols %>%
group_by(sample, marker_id) %>%
summarise(n_haplotypes = n(), .groups = 'drop') %>%
group_by(sample) %>%
summarise(max_moi = max(n_haplotypes)) ## # A tibble: 135 × 2
## sample max_moi
## <chr> <int>
## 1 AR-002-0 1
## 2 AR-006-0 2
## 3 AR-007-0 1
## 4 AR-008-0 2
## 5 AR-009-0 2
## 6 AR-015-0 2
## 7 AR-023-0 2
## 8 AR-024-0 1
## 9 AR-024-112 2
## 10 AR-024-56 2
## # ℹ 125 more rowsRecurrences
Participants with more than one episode for inference (n=41 participants, n=99 Pv isolates):
sols %>%
select(info, episodes) %>%
distinct() %>%
arrange(info) %>%
filter(episodes>1) ## info episodes
## 1 AR-024 5
## 2 AR-067 2
## 3 AR-081 3
## 4 AR-098 2
## 5 AR-110 2
## 6 AR-133 3
## 7 AR-140 3
## 8 AR-142 4
## 9 AR-160 2
## 10 AR-173 2
## 11 AR-176 3
## 12 AR-187 4
## 13 AR-192 2
## 14 AR-214 2
## 15 AR-221 2
## 16 AR-234 2
## 17 AR-237 2
## 18 AR-246 2
## 19 AR-248 2
## 20 AR-250 2
## 21 AR-258 2
## 22 AR-274 2
## 23 AR-288 2
## 24 AR-300 2
## 25 AR-302 2
## 26 AR-304 2
## 27 AR-306 3
## 28 AR-322 2
## 29 AR-325 2
## 30 AR-326 2
## 31 AR-328 3
## 32 AR-341 2
## 33 AR-343 2
## 34 AR-345 2
## 35 AR-346 2
## 36 AR-348 2
## 37 AR-355 2
## 38 AR-359 6
## 39 AR-366 2
## 40 AR-382 2
## 41 AR-383 2ids_recurrent <- sols %>% clean_names() %>% filter(episodes > 1) %>% distinct(info)
sols_recurrent <- sols %>% clean_names() %>%filter(episodes > 1)
# setdiff(ids_recurrent$info, sols_recurrent$info)
# setdiff(sols_recurrent$info, ids_recurrent$info)
# sols_recurrent %>% distinct(sample) %>% arrange(sample) # 99 isolatessols_recurrent %>%
distinct(sample) %>%
separate(sample, into = c("sample", "episode_day"), sep = "-(?=[0-9]+$)") %>%
ggplot(aes(x = episode_day, y = sample)) +
geom_line(aes(group = sample), color = "darkgrey") +
geom_point() +
theme_bw()
Data on episode number and time since last episode
episode_summary <- sols_recurrent %>%
distinct(info, info_d, vis_date, sample) %>%
mutate(vis_date = mdy(vis_date)) %>%
arrange(info, vis_date) %>%
group_by(info) %>%
mutate(
# get the episode number
episode_number = row_number(),
# calculate days since enrolment, just a check with dates
days_since_enrolment = as.integer(vis_date - min(vis_date[info_d == 0])),
# calculate days since last episode using lag() and making enrolment episodes 0 days since last
days_since_last_episode = replace_na(as.integer(vis_date - lag(vis_date)), 0)
) %>%
ungroup()Haplotype frequencies
We want to remove any haplotypes that appear only once at very low frequencies. Haplotypes should already be filtered to be observed in at least 2 samples and within-host frequency >=1% (in full dataset).
singleton_haps <- sols_recurrent %>%
select(sample, marker_id, haplotype, frequency, count) %>%
count(haplotype) %>%
arrange(n) %>%
filter(n==1) %>%
pull(haplotype)
sols_recurrent %>%
filter(haplotype %in% singleton_haps) %>%
ggplot(aes(x = haplotype, y = count)) +
geom_hline(yintercept = 100, linetype = "dashed") +
geom_point(aes(color = frequency, shape = follow_x), size = 3) +
labs(x = "singleton haplotype",
y = "read count",
color = "within-sample frequency",
shape = "timepoint") +
theme_bw() +
theme(axis.text.x = element_text(angle = 90, hjust = 1, vjust = 0.5))
## Haplotypes for inference As per Aimee, to avoid bias due to
within-host selection of recrudescent parasites, we use only enrolment
episodes to estimate population-level alelle frequencies. So, we will
filter out any haplotypes that do not appear at baseline (drop n=8
haplotypes). Details below on their within-sample frequencies and read
counts.
# n=569 haps
baseline_haps <- sols_recurrent %>%
filter(follow_x == "baseline") %>%
pull(haplotype)
sols_recurrent %>%
filter(!haplotype %in% baseline_haps) %>%
select(sample, marker_id, haplotype, frequency, count) %>%
arrange(frequency, count)## sample marker_id haplotype frequency count
## 1 AR-081-168 Chr09 Chr09-7 0.1066 716
## 2 AR-343-112 Chr07 Chr07-11 0.2332 2302
## 3 AR-081-140 Chr09 Chr09-7 0.7282 3518
## 4 AR-024-70 Chr05 Chr05-13 0.9918 7028
## 5 AR-024-56 Chr05 Chr05-13 0.9933 4720
## 6 AR-081-168 Chr05 Chr05-10 1.0000 9622
## 7 AR-024-84 Chr05 Chr05-13 1.0000 9693
## 8 AR-081-140 Chr05 Chr05-10 1.0000 9745Final data for inference
analysis_data <- sols_recurrent %>%
select(subject_id = info,
sample_id = sample,
treatment_arm = trt_label,
days_since_treatment = info_d,
timepoint = follow_x,
visit_date = vis_date,
age_years = age_y,
sex = gender,
episodes,
marker_id,
haplotype,
type,
frequency,
mean_moi,
max_moi) %>%
# add extra epi info on episode number and time since last episode
left_join(episode_summary %>% select(subject_id = info,
sample_id = sample,
episode_number,
days_since_enrolment,
days_since_last_episode),
by = c("subject_id", "sample_id")) %>%
# keep only haplotypes present at baseline
filter(haplotype %in% baseline_haps) %>%
# ensure dates are date class
mutate(visit_date = mdy(visit_date)) Baseline allele frequencies
test_df <- analysis_data %>%
# Use only baseline samples
filter(timepoint == "baseline") %>%
# filter data frame by marker
filter(marker_id == "Chr11")
test_df %>%
mutate(haplotype = factor(haplotype)) %>%
select(sample_id, haplotype, frequency) %>%
pivot_wider(names_from = haplotype,
values_from = frequency,
values_fill = 0) %>%
pivot_longer(cols = -sample_id,
names_to = "haplotype",
values_to = "frequency") %>%
# Get population-level haplotype frequency,
# correcting for when within-individual sum is not equal
# to 1, as can happen when a minority clone is <2%
group_by(sample_id) %>%
mutate(frequency = frequency / sum(frequency, na.rm = TRUE)) %>%
# get population-level mean freq
group_by(haplotype) %>%
summarise(poplevel_mean_freq = mean(frequency, na.rm = TRUE)) %>%
adorn_totals("row")## Derive baseline allele frequency - This is modified from Thomas/Jason script for our data
baseline_fs <- analysis_data %>%
# Use only baseline samples
filter(timepoint == "baseline") %>%
# split data frame by marker
group_by(marker_id) %>%
group_split() %>%
# Derive a within-marker list of frequencies, by individual
lapply(function(x) {
x <- x %>%
# Build a within-individual frequency table that
# always includes every haplotype (even the ones absent)
mutate(haplotype = factor(haplotype)) %>%
select(sample_id, haplotype, frequency) %>%
pivot_wider(names_from = haplotype,
values_from = frequency,
values_fill = 0) %>%
pivot_longer(cols = -sample_id,
names_to = "haplotype",
values_to = "frequency") %>%
# Get population-level haplotype frequency,
# correcting for when within-individual sum is not equal
# to 1, as can happen when a minority clone is <2%
group_by(sample_id) %>%
mutate(frequency = frequency / sum(frequency, na.rm = TRUE)) %>%
group_by(haplotype) %>%
summarise(frequency_pop_mean = mean(frequency, na.rm = TRUE))
return(deframe(x))
}) %>%
# get marker_id from group_keys from the group dfs
setNames(nm = analysis_data %>% group_by(marker_id) %>% group_keys() %>% pull(marker_id))
# Here we can also save as dataframe for easier printing and table-ready for paper
baseline_fs_df <- analysis_data %>%
# Use only baseline samples
filter(timepoint == "baseline") %>%
# Group by marker_id and sample_id for further calculations
group_by(marker_id, sample_id) %>%
# Build a within-individual frequency table that always includes every haplotype (even the ones absent)
mutate(haplotype = factor(haplotype)) %>%
select(marker_id, sample_id, haplotype, frequency) %>%
pivot_wider(names_from = haplotype, values_from = frequency, values_fill = list(frequency = 0)) %>%
pivot_longer(cols = -c(marker_id, sample_id), names_to = "haplotype", values_to = "frequency") %>%
# Get population-level haplotype frequency, correcting for when within-individual sum is not equal to 1
group_by(marker_id, sample_id) %>%
mutate(frequency = frequency / sum(frequency, na.rm = TRUE)) %>%
group_by(marker_id, haplotype) %>%
summarise(frequency_pop_mean = mean(frequency, na.rm = TRUE), .groups = 'drop') %>%
# Ensure haplotypes are correctly associated with their marker_id - note that this works for us because , in future would have to make this flexible to allow for haplotype names that are not reliant on having marker_id
filter(str_detect(haplotype, marker_id))
baseline_fs_df## # A tibble: 67 × 3
## marker_id haplotype frequency_pop_mean
## <chr> <chr> <dbl>
## 1 Chr01 Chr01-1 0.534
## 2 Chr01 Chr01-2 0.330
## 3 Chr01 Chr01-3 0.112
## 4 Chr01 Chr01-4 0.0244
## 5 Chr02 Chr02-1 0.228
## 6 Chr02 Chr02-11 0.00467
## 7 Chr02 Chr02-2 0.133
## 8 Chr02 Chr02-3 0.198
## 9 Chr02 Chr02-4 0.191
## 10 Chr02 Chr02-5 0.101
## # ℹ 57 more rowsPv3Rs
Example on one participant: AR-067
# This has been modified from Thomas/Jason script
## Pick an individual at random to run Pv3Rs
# indiv_name <- sample(unique(analysis_data$subject_id), size = 1)
indiv_name <- "AR-067"
## Prepare the data
# 1- Subset haplotype data to specific individual and apply filters
indiv_haplotype_data <- analysis_data %>%
# Restrict to a single patient
filter(subject_id == indiv_name) %>%
# Restrict to a subset of markers, if needed
filter(marker_id %in% BENCHMARK_MARKERS) %>%
# Restrict to summed MOI below threshold
group_by(subject_id, episode_number, marker_id) %>%
mutate(MOI_per_marker = sum(n())) %>%
group_by(subject_id, episode_number) %>%
mutate(MOI_per_episode = max(MOI_per_marker, na.rm = TRUE)) %>%
ungroup() %>%
mutate(marker_id = factor(marker_id, levels = BENCHMARK_MARKERS))
# 2- Calculate per-episode and per-participant MOI for PvR3S eligibility
indiv_MOI <- indiv_haplotype_data %>%
select(subject_id, episode_number,
marker_id, starts_with("MOI_")) %>%
distinct() %>%
group_by(subject_id, episode_number) %>%
# Get highest per-marker MOI only for each episode
# (drop marker_id in case of ties with highest per-marker MOI)
select(-marker_id) %>%
distinct() %>%
filter(MOI_per_marker == max(MOI_per_marker, na.rm = TRUE)) %>%
ungroup() %>%
mutate(MOI_summed = sum(MOI_per_episode, na.rm = TRUE))Prepare test data
# Finish data preparation
indiv_haplotype_data <- indiv_haplotype_data %>%
group_by(episode_number) %>%
group_split() %>%
lapply(function(x) {
res <- x %>%
select(sample_id, episode_number,
marker_id, haplotype, frequency) %>%
# For sensitivity analysis, allow to include/drop allele
# based on their within-individual frequency
filter(frequency >= WITHIN_INDIVIDUAL_ALLELE_FREQ_THR) %>%
select(-sample_id, -episode_number, -frequency) %>%
distinct() %>%
# Prevent dropping of markers that are not characterised
# by setting .drop to FALSE
group_by(marker_id, .drop = FALSE) %>%
group_split() %>%
lapply(function(y) {
unique(y$haplotype)
})
# Returned a list named with each episode,
# setting marker allele to NA in case none are observed
return(lapply(setNames(res, BENCHMARK_MARKERS),
function(y) {
if (length(y) == 0) return(NA) else return(y)
}))
})# Run Aimee's posterior estimation
indiv_posterior <- Pv3Rs::compute_posterior(y = indiv_haplotype_data,
fs = baseline_fs[BENCHMARK_MARKERS])## Number of valid relationship graphs (RGs) is 39## =============================================================================|
## Computing log p(Y|RG) for 39 RGs
## =============================================================================|
## Finding log-likelihood of each vector of recurrent states
## =============================================================================|indiv_posterior## $marg
## C L I
## [1,] 0 0.9971211 0.00287892
##
## $joint
## C L I
## 0.00000000 0.99712108 0.00287892The joint posterior probability of relapse is 99.7%, when we look at the genetic data, this is in line with the inferred probability of relpase.
sols_recurrent %>%
filter(info == "AR-067") %>%
ggplot(aes(x = haplotype, y = factor(info_d), fill = factor(haplotype))) +
geom_tile() +
facet_grid(~marker_id, scales = "free_x") +
theme_bw() +
theme(axis.text.x = element_text(angle = 90, hjust = 1, vjust = 0.5))
### Run Pv3Rs one random individual
# This has been modified from Thomas/Jason script
## Pick an individual at random to run Pv3Rs
indiv_name <- sample(unique(analysis_data$subject_id), size = 1)
## Prepare the data
# 1- Subset haplotype data to specific individual and apply filters
indiv_haplotype_data <- analysis_data %>%
# Restrict to a single patient
filter(subject_id == indiv_name) %>%
# Restrict to a subset of markers, if needed
filter(marker_id %in% BENCHMARK_MARKERS) %>%
# Restrict to summed MOI below threshold
group_by(subject_id, episode_number, marker_id) %>%
mutate(MOI_per_marker = sum(n())) %>%
group_by(subject_id, episode_number) %>%
mutate(MOI_per_episode = max(MOI_per_marker, na.rm = TRUE)) %>%
ungroup() %>%
mutate(marker_id = factor(marker_id, levels = BENCHMARK_MARKERS))
# 2- Calculate per-episode and per-participant MOI for PvR3S eligibility
indiv_MOI <- indiv_haplotype_data %>%
select(subject_id, episode_number,
marker_id, starts_with("MOI_")) %>%
distinct() %>%
group_by(subject_id, episode_number) %>%
# Get highest per-marker MOI only for each episode
# (drop marker_id in case of ties with highest per-marker MOI)
select(-marker_id) %>%
distinct() %>%
filter(MOI_per_marker == max(MOI_per_marker, na.rm = TRUE)) %>%
ungroup() %>%
mutate(MOI_summed = sum(MOI_per_episode, na.rm = TRUE))Now that the data has been subset, Pv3Rs::compute_posterior() will be called (only if the participant meets eligibility criteria as defined by the global variables set earlier in this report).
if (RUN_EXAMPLE & unique(indiv_MOI$MOI_summed) <= MAX_MOI_TO_INCLUDE) {
# Verbose
cat("Running Pv3Rs for: ",
indiv_name,
" (summed MOI = ",
unique(indiv_MOI$MOI_summed),
").\n",
sep = "")
# Finish data preparation
indiv_haplotype_data <- indiv_haplotype_data %>%
group_by(episode_number) %>%
group_split() %>%
lapply(function(x) {
res <- x %>%
select(sample_id, episode_number,
marker_id, haplotype, frequency) %>%
# For sensitivity analysis, allow to include/drop allele
# based on their within-individual frequency
filter(frequency >= WITHIN_INDIVIDUAL_ALLELE_FREQ_THR) %>%
select(-sample_id, -episode_number, -frequency) %>%
distinct() %>%
# Prevent dropping of markers that are not characterised
# by setting .drop to FALSE
group_by(marker_id, .drop = FALSE) %>%
group_split() %>%
lapply(function(y) {
unique(y$haplotype)
})
# Returned a list named with each episode,
# setting marker allele to NA in case none are observed
return(lapply(setNames(res, BENCHMARK_MARKERS),
function(y) {
if (length(y) == 0) return(NA) else return(y)
}))
})
# Run Aimee's posterior estimation
indiv_posterior <- Pv3Rs::compute_posterior(y = indiv_haplotype_data,
fs = baseline_fs[BENCHMARK_MARKERS])
} else {
# Verbose
cat("NOT Running Pv3Rs for: ",
indiv_name,
" because either summed MOI exceeds threshold (observed = ",
unique(indiv_MOI$MOI_summed),
", MAX_MOI_TO_INCLUDE = ",
MAX_MOI_TO_INCLUDE,
"), or RUN_EXAMPLE was set to FALSE.\n",
sep = "")
# Return NULL
indiv_posterior <- NULL
}
indiv_posteriorsols_recurrent %>%
filter(info == indiv_name) %>%
ggplot(aes(x = haplotype, y = factor(info_d), fill = factor(haplotype))) +
geom_tile() +
facet_grid(~marker_id, scales = "free_x") +
theme_bw() +
theme(axis.text.x = element_text(angle = 90, hjust = 1, vjust = 0.5))Run for all participants
# Start timer
t_start <- Sys.time()
# Initialize an empty list to store the results
indiv_posteriors <- list()
# Loop through each unique individual name
for (indiv_name in unique(analysis_data$subject_id)) {
# Verbose
cat("ID : ", indiv_name, "...\n", sep = "")
## Prepare the data
# 1- Subset haplotype data to specific individual and apply filters
indiv_haplotype_data <- analysis_data %>%
# Restrict to a single patient
filter(subject_id == indiv_name) %>%
# Ensure episodes are in order
arrange(episode_number) %>%
# Restrict to a subset of markers
filter(marker_id %in% BENCHMARK_MARKERS) %>%
# Restrict to summed MOI below threshold
group_by(subject_id, episode_number, marker_id) %>%
mutate(MOI_per_marker = sum(n())) %>%
group_by(subject_id, episode_number) %>%
mutate(MOI_per_episode = max(MOI_per_marker, na.rm = TRUE)) %>%
ungroup() %>%
mutate(marker_id = factor(marker_id, levels = BENCHMARK_MARKERS))
# 2- Calculate per-episode and per-participant MOI for PvR3S eligibility
indiv_MOI <- indiv_haplotype_data %>%
select(subject_id, episode_number,
marker_id, starts_with("MOI_")) %>%
distinct() %>%
group_by(subject_id, episode_number) %>%
# Get highest per-marker MOI only for each episode
# (drop marker_id in case of ties with highest per-marker MOI)
select(-marker_id) %>%
distinct() %>%
filter(MOI_per_marker == max(MOI_per_marker, na.rm = TRUE)) %>%
ungroup() %>%
mutate(MOI_summed = sum(MOI_per_episode, na.rm = TRUE))
## Run Pv3Rs only if MOI below threshold
if (unique(indiv_MOI$MOI_summed) <= MAX_MOI_TO_INCLUDE) {
# Verbose
cat("Running Pv3Rs for: ",
indiv_name,
" (summed MOI = ",
unique(indiv_MOI$MOI_summed),
").\n",
sep = "")
# Preserve episode numbers for naming the output list
indiv_episode <- unique(indiv_haplotype_data$episode_number)
cat("The episode number is: ", indiv_episode) # printing to check episode order number is correct
# Finish data preparation
indiv_haplotype_data <- indiv_haplotype_data %>%
group_by(episode_number) %>%
group_split() %>%
lapply(function(x) {
res <- x %>%
select(sample_id, episode_number,
marker_id, haplotype, frequency) %>%
# For sensitivity analysis, allow to include/drop allele
# based on their within-individual frequency
filter(frequency >= WITHIN_INDIVIDUAL_ALLELE_FREQ_THR) %>%
select(-sample_id, -episode_number, -frequency) %>%
distinct() %>%
# Prevent dropping of markers that are not characterized
# by setting .drop to FALSE
group_by(marker_id, .drop = FALSE) %>%
group_split() %>%
lapply(function(y) {
unique(y$haplotype)
})
# Returned a list named with each episode,
# setting marker allele to NA in case none are observed
return(lapply(setNames(res, BENCHMARK_MARKERS),
function(y) {
if (length(y) == 0) return(NA) else return(y)
}))
})
# Run Aimee's posterior estimation
indiv_posterior <- compute_posterior(y = indiv_haplotype_data,
fs = baseline_fs[BENCHMARK_MARKERS],
prior = matrix(PRIOR_3RS,
nrow = length(indiv_haplotype_data),
ncol = length(PRIOR_3RS),
byrow = TRUE,
dimnames = list(c(1:length(indiv_haplotype_data)),
names(PRIOR_3RS))))
} else {
# Verbose
cat("NOT Running Pv3Rs for: ",
indiv_name,
" because summed MOI exceeds threshold (observed = ",
unique(indiv_MOI$MOI_summed),
", MAX_MOI_TO_INCLUDE = ",
MAX_MOI_TO_INCLUDE,
").\n",
sep = "")
# Return NULL
indiv_episode <- unique(indiv_haplotype_data$episode_number)
indiv_posterior <- NULL
}
# Append the results to the list
indiv_posteriors[[indiv_name]] <- list("subject_id" = indiv_name,
"episode_number" = indiv_episode,
"Pv3Rs" = indiv_posterior)
}## ID : AR-024...
## Running Pv3Rs for: AR-024 (summed MOI = 7).
## The episode number is: 1 2 3 4 5## Number of valid relationship graphs (RGs) is 14718## =============================================================================|
## Computing log p(Y|RG) for 14718 RGs
## =============================================================================|
## Finding log-likelihood of each vector of recurrent states
## =============================================================================|## ID : AR-221...
## Running Pv3Rs for: AR-221 (summed MOI = 3).
## The episode number is: 1 2## Number of valid relationship graphs (RGs) is 9
## =============================================================================|
## Computing log p(Y|RG) for 9 RGs
## =============================================================================|
## Finding log-likelihood of each vector of recurrent states
## =============================================================================|## ID : AR-343...
## Running Pv3Rs for: AR-343 (summed MOI = 4).
## The episode number is: 1 2## Number of valid relationship graphs (RGs) is 39
## =============================================================================|
## Computing log p(Y|RG) for 39 RGs
## =============================================================================|
## Finding log-likelihood of each vector of recurrent states
## =============================================================================|## ID : AR-366...
## Running Pv3Rs for: AR-366 (summed MOI = 4).
## The episode number is: 1 2## Number of valid relationship graphs (RGs) is 30
## =============================================================================|
## Computing log p(Y|RG) for 30 RGs
## =============================================================================|
## Finding log-likelihood of each vector of recurrent states
## =============================================================================|## ID : AR-187...
## Running Pv3Rs for: AR-187 (summed MOI = 5).
## The episode number is: 1 2 3 4## Number of valid relationship graphs (RGs) is 298
## =============================================================================|
## Computing log p(Y|RG) for 298 RGs
## =============================================================================|
## Finding log-likelihood of each vector of recurrent states
## =============================================================================|## ID : AR-383...
## Running Pv3Rs for: AR-383 (summed MOI = 2).
## The episode number is: 1 2## Number of valid relationship graphs (RGs) is 3
## =============================================================================|
## Computing log p(Y|RG) for 3 RGs
## =============================================================================|
## Finding log-likelihood of each vector of recurrent states
## =============================================================================|## ID : AR-288...
## Running Pv3Rs for: AR-288 (summed MOI = 4).
## The episode number is: 1 2## Number of valid relationship graphs (RGs) is 39
## =============================================================================|
## Computing log p(Y|RG) for 39 RGs
## =============================================================================|
## Finding log-likelihood of each vector of recurrent states
## =============================================================================|## ID : AR-234...
## Running Pv3Rs for: AR-234 (summed MOI = 5).
## The episode number is: 1 2## Number of valid relationship graphs (RGs) is 172
## =============================================================================|
## Computing log p(Y|RG) for 172 RGs
## =============================================================================|
## Finding log-likelihood of each vector of recurrent states
## =============================================================================|## ID : AR-306...
## Running Pv3Rs for: AR-306 (summed MOI = 5).
## The episode number is: 1 2 3## Number of valid relationship graphs (RGs) is 250
## =============================================================================|
## Computing log p(Y|RG) for 250 RGs
## =============================================================================|
## Finding log-likelihood of each vector of recurrent states
## =============================================================================|## ID : AR-302...
## Running Pv3Rs for: AR-302 (summed MOI = 4).
## The episode number is: 1 2## Number of valid relationship graphs (RGs) is 39
## =============================================================================|
## Computing log p(Y|RG) for 39 RGs
## =============================================================================|
## Finding log-likelihood of each vector of recurrent states
## =============================================================================|## ID : AR-341...
## Running Pv3Rs for: AR-341 (summed MOI = 5).
## The episode number is: 1 2## Number of valid relationship graphs (RGs) is 172
## =============================================================================|
## Computing log p(Y|RG) for 172 RGs
## =============================================================================|
## Finding log-likelihood of each vector of recurrent states
## =============================================================================|## ID : AR-326...
## Running Pv3Rs for: AR-326 (summed MOI = 5).
## The episode number is: 1 2## Number of valid relationship graphs (RGs) is 172
## =============================================================================|
## Computing log p(Y|RG) for 172 RGs
## =============================================================================|
## Finding log-likelihood of each vector of recurrent states
## =============================================================================|## ID : AR-067...
## Running Pv3Rs for: AR-067 (summed MOI = 4).
## The episode number is: 1 2## Number of valid relationship graphs (RGs) is 39
## =============================================================================|
## Computing log p(Y|RG) for 39 RGs
## =============================================================================|
## Finding log-likelihood of each vector of recurrent states
## =============================================================================|## ID : AR-322...
## Running Pv3Rs for: AR-322 (summed MOI = 3).
## The episode number is: 1 2## Number of valid relationship graphs (RGs) is 9
## =============================================================================|
## Computing log p(Y|RG) for 9 RGs
## =============================================================================|
## Finding log-likelihood of each vector of recurrent states
## =============================================================================|## ID : AR-355...
## Running Pv3Rs for: AR-355 (summed MOI = 3).
## The episode number is: 1 2## Number of valid relationship graphs (RGs) is 9
## =============================================================================|
## Computing log p(Y|RG) for 9 RGs
## =============================================================================|
## Finding log-likelihood of each vector of recurrent states
## =============================================================================|## ID : AR-081...
## Running Pv3Rs for: AR-081 (summed MOI = 7).
## The episode number is: 1 2 3## Number of valid relationship graphs (RGs) is 9773
## =============================================================================|
## Computing log p(Y|RG) for 9773 RGs
## =============================================================================|
## Finding log-likelihood of each vector of recurrent states
## =============================================================================|## ID : AR-248...
## Running Pv3Rs for: AR-248 (summed MOI = 4).
## The episode number is: 1 2## Number of valid relationship graphs (RGs) is 39
## =============================================================================|
## Computing log p(Y|RG) for 39 RGs
## =============================================================================|
## Finding log-likelihood of each vector of recurrent states
## =============================================================================|## ID : AR-304...
## Running Pv3Rs for: AR-304 (summed MOI = 5).
## The episode number is: 1 2## Number of valid relationship graphs (RGs) is 172
## =============================================================================|
## Computing log p(Y|RG) for 172 RGs
## =============================================================================|
## Finding log-likelihood of each vector of recurrent states
## =============================================================================|## ID : AR-160...
## Running Pv3Rs for: AR-160 (summed MOI = 2).
## The episode number is: 1 2## Number of valid relationship graphs (RGs) is 3
## =============================================================================|
## Computing log p(Y|RG) for 3 RGs
## =============================================================================|
## Finding log-likelihood of each vector of recurrent states
## =============================================================================|## ID : AR-110...
## Running Pv3Rs for: AR-110 (summed MOI = 2).
## The episode number is: 1 2## Number of valid relationship graphs (RGs) is 3
## =============================================================================|
## Computing log p(Y|RG) for 3 RGs
## =============================================================================|
## Finding log-likelihood of each vector of recurrent states
## =============================================================================|## ID : AR-133...
## Running Pv3Rs for: AR-133 (summed MOI = 5).
## The episode number is: 1 2 3## Number of valid relationship graphs (RGs) is 202
## =============================================================================|
## Computing log p(Y|RG) for 202 RGs
## =============================================================================|
## Finding log-likelihood of each vector of recurrent states
## =============================================================================|## ID : AR-140...
## Running Pv3Rs for: AR-140 (summed MOI = 7).
## The episode number is: 1 2 3## Number of valid relationship graphs (RGs) is 9773
## =============================================================================|
## Computing log p(Y|RG) for 9773 RGs
## =============================================================================|
## Finding log-likelihood of each vector of recurrent states
## =============================================================================|## ID : AR-142...
## Running Pv3Rs for: AR-142 (summed MOI = 8).
## The episode number is: 1 2 3 4## Number of valid relationship graphs (RGs) is 91237
## =============================================================================|
## Computing log p(Y|RG) for 91237 RGs
## =============================================================================|
## Finding log-likelihood of each vector of recurrent states
## =============================================================================|## ID : AR-173...
## Running Pv3Rs for: AR-173 (summed MOI = 3).
## The episode number is: 1 2## Number of valid relationship graphs (RGs) is 9
## =============================================================================|
## Computing log p(Y|RG) for 9 RGs
## =============================================================================|
## Finding log-likelihood of each vector of recurrent states
## =============================================================================|## ID : AR-328...
## Running Pv3Rs for: AR-328 (summed MOI = 5).
## The episode number is: 1 2 3## Number of valid relationship graphs (RGs) is 250
## =============================================================================|
## Computing log p(Y|RG) for 250 RGs
## =============================================================================|
## Finding log-likelihood of each vector of recurrent states
## =============================================================================|## ID : AR-359...
## NOT Running Pv3Rs for: AR-359 because summed MOI exceeds threshold (observed = 11, MAX_MOI_TO_INCLUDE = 8).
## ID : AR-325...
## Running Pv3Rs for: AR-325 (summed MOI = 2).
## The episode number is: 1 2## Number of valid relationship graphs (RGs) is 3
## =============================================================================|
## Computing log p(Y|RG) for 3 RGs
## =============================================================================|
## Finding log-likelihood of each vector of recurrent states
## =============================================================================|## ID : AR-246...
## Running Pv3Rs for: AR-246 (summed MOI = 2).
## The episode number is: 1 2## Number of valid relationship graphs (RGs) is 3
## =============================================================================|
## Computing log p(Y|RG) for 3 RGs
## =============================================================================|
## Finding log-likelihood of each vector of recurrent states
## =============================================================================|## ID : AR-250...
## Running Pv3Rs for: AR-250 (summed MOI = 3).
## The episode number is: 1 2## Number of valid relationship graphs (RGs) is 9
## =============================================================================|
## Computing log p(Y|RG) for 9 RGs
## =============================================================================|
## Finding log-likelihood of each vector of recurrent states
## =============================================================================|## ID : AR-258...
## Running Pv3Rs for: AR-258 (summed MOI = 5).
## The episode number is: 1 2## Number of valid relationship graphs (RGs) is 172
## =============================================================================|
## Computing log p(Y|RG) for 172 RGs
## =============================================================================|
## Finding log-likelihood of each vector of recurrent states
## =============================================================================|## ID : AR-098...
## Running Pv3Rs for: AR-098 (summed MOI = 3).
## The episode number is: 1 2## Number of valid relationship graphs (RGs) is 9
## =============================================================================|
## Computing log p(Y|RG) for 9 RGs
## =============================================================================|
## Finding log-likelihood of each vector of recurrent states
## =============================================================================|## ID : AR-176...
## Running Pv3Rs for: AR-176 (summed MOI = 5).
## The episode number is: 1 2 3## Number of valid relationship graphs (RGs) is 250
## =============================================================================|
## Computing log p(Y|RG) for 250 RGs
## =============================================================================|
## Finding log-likelihood of each vector of recurrent states
## =============================================================================|## ID : AR-192...
## Running Pv3Rs for: AR-192 (summed MOI = 3).
## The episode number is: 1 2## Number of valid relationship graphs (RGs) is 9
## =============================================================================|
## Computing log p(Y|RG) for 9 RGs
## =============================================================================|
## Finding log-likelihood of each vector of recurrent states
## =============================================================================|## ID : AR-214...
## Running Pv3Rs for: AR-214 (summed MOI = 6).
## The episode number is: 1 2## Number of valid relationship graphs (RGs) is 799
## =============================================================================|
## Computing log p(Y|RG) for 799 RGs
## =============================================================================|
## Finding log-likelihood of each vector of recurrent states
## =============================================================================|## ID : AR-237...
## Running Pv3Rs for: AR-237 (summed MOI = 3).
## The episode number is: 1 2## Number of valid relationship graphs (RGs) is 9
## =============================================================================|
## Computing log p(Y|RG) for 9 RGs
## =============================================================================|
## Finding log-likelihood of each vector of recurrent states
## =============================================================================|## ID : AR-274...
## Running Pv3Rs for: AR-274 (summed MOI = 4).
## The episode number is: 1 2## Number of valid relationship graphs (RGs) is 39
## =============================================================================|
## Computing log p(Y|RG) for 39 RGs
## =============================================================================|
## Finding log-likelihood of each vector of recurrent states
## =============================================================================|## ID : AR-300...
## Running Pv3Rs for: AR-300 (summed MOI = 4).
## The episode number is: 1 2## Number of valid relationship graphs (RGs) is 39
## =============================================================================|
## Computing log p(Y|RG) for 39 RGs
## =============================================================================|
## Finding log-likelihood of each vector of recurrent states
## =============================================================================|## ID : AR-345...
## Running Pv3Rs for: AR-345 (summed MOI = 4).
## The episode number is: 1 2## Number of valid relationship graphs (RGs) is 39
## =============================================================================|
## Computing log p(Y|RG) for 39 RGs
## =============================================================================|
## Finding log-likelihood of each vector of recurrent states
## =============================================================================|## ID : AR-346...
## Running Pv3Rs for: AR-346 (summed MOI = 5).
## The episode number is: 1 2## Number of valid relationship graphs (RGs) is 172
## =============================================================================|
## Computing log p(Y|RG) for 172 RGs
## =============================================================================|
## Finding log-likelihood of each vector of recurrent states
## =============================================================================|## ID : AR-348...
## Running Pv3Rs for: AR-348 (summed MOI = 2).
## The episode number is: 1 2## Number of valid relationship graphs (RGs) is 3
## =============================================================================|
## Computing log p(Y|RG) for 3 RGs
## =============================================================================|
## Finding log-likelihood of each vector of recurrent states
## =============================================================================|## ID : AR-382...
## Running Pv3Rs for: AR-382 (summed MOI = 2).
## The episode number is: 1 2## Number of valid relationship graphs (RGs) is 3
## =============================================================================|
## Computing log p(Y|RG) for 3 RGs
## =============================================================================|
## Finding log-likelihood of each vector of recurrent states
## =============================================================================|# End timer
t_end <- Sys.time()
cat("Pv3Rs for the whole dataset took : ", as.numeric(difftime(time1 = t_end,
time2 = t_start,
units = "secs"))/60, " mins", "\n", sep = "")## Pv3Rs for the whole dataset took : 1.561027 mins# Present the marginal data in a clearer format
indiv_posteriors_marginal <- do.call(rbind,
lapply(indiv_posteriors, function(x) {
if (!is.null(x[["Pv3Rs"]])) {
return(data.frame("subject_id" = x[["subject_id"]],
"episode_number" = x[["episode_number"]][-1],
"Posterior_marginal_prob_C" = x[["Pv3Rs"]]$marg[, "C"],
"Posterior_marginal_prob_L" = x[["Pv3Rs"]]$marg[, "L"],
"Posterior_marginal_prob_I" = x[["Pv3Rs"]]$marg[, "I"]))
} else {
return(NULL)
}
}))
row.names(indiv_posteriors_marginal) <- 1:nrow(indiv_posteriors_marginal)
# Present the joint posterior estimates in a clearer format
indiv_posteriors_joint <- do.call(rbind,
lapply(indiv_posteriors, function(x) {
if (!is.null(x[["Pv3Rs"]])) {
joint_probs <- x[["Pv3Rs"]]$joint
# Extract the state pairs and probabilities
state_pairs <- names(joint_probs)
prob_values <- as.numeric(joint_probs)
# Create a data frame with subject_id, episode_number, and joint probabilities
return(data.frame("subject_id" = rep(x[["subject_id"]], length(state_pairs)),
"episode_number" = rep(x[["episode_number"]][-1], each = length(state_pairs)),
"state_pair" = state_pairs,
"joint_probability" = prob_values))
} else {
return(NULL)
}
}))
row.names(indiv_posteriors_joint) <- 1:nrow(indiv_posteriors_joint)
# Save Pv3Rs output because it's time consuming and
# we don't want to re-run it every time.
save(list = c("RUN_EXAMPLE", "WITHIN_INDIVIDUAL_ALLELE_FREQ_THR", "BENCHMARK_MARKERS", "MAX_MOI_TO_INCLUDE", "PRIOR_3RS",
"analysis_data", "baseline_fs",
"indiv_posteriors", "indiv_posteriors_marginal", "indiv_posteriors_joint"),
file = paste0("./outputs/Pv3Rs_sols_posteriors_",
strftime(Sys.time(), format = "%Y%m%d_%H%M%S"),
".RData"))Explore the posterior
The marginal probabilities give us the probability of the three states for each recurrent episode. However, this does not consider the joint probability of different states when a person experienced more than one recurrent episode.
indiv_posteriors_marginal## subject_id episode_number Posterior_marginal_prob_C
## 1 AR-024 2 0.0000000
## 2 AR-024 3 0.7576302
## 3 AR-024 4 0.8623076
## 4 AR-024 5 0.0000000
## 5 AR-221 2 0.0000000
## 6 AR-343 2 0.0000000
## 7 AR-366 2 0.0000000
## 8 AR-187 2 0.7721759
## 9 AR-187 3 0.7721759
## 10 AR-187 4 0.0000000
## 11 AR-383 2 0.0000000
## 12 AR-288 2 0.0000000
## 13 AR-234 2 0.0000000
## 14 AR-306 2 0.0000000
## 15 AR-306 3 0.7913832
## 16 AR-302 2 0.0000000
## 17 AR-341 2 0.0000000
## 18 AR-326 2 0.0000000
## 19 AR-067 2 0.0000000
## 20 AR-322 2 0.0000000
## 21 AR-355 2 0.0000000
## 22 AR-081 2 0.0000000
## 23 AR-081 3 0.0000000
## 24 AR-248 2 0.0000000
## 25 AR-304 2 0.0000000
## 26 AR-160 2 0.7485703
## 27 AR-110 2 0.7479272
## 28 AR-133 2 0.8615406
## 29 AR-133 3 0.0000000
## 30 AR-140 2 0.0000000
## 31 AR-140 3 0.0000000
## 32 AR-142 2 0.0000000
## 33 AR-142 3 0.0000000
## 34 AR-142 4 0.0000000
## 35 AR-173 2 0.0000000
## 36 AR-328 2 0.0000000
## 37 AR-328 3 0.0000000
## 38 AR-325 2 0.0000000
## 39 AR-246 2 0.0000000
## 40 AR-250 2 0.6878086
## 41 AR-258 2 0.0000000
## 42 AR-098 2 0.0000000
## 43 AR-176 2 0.0000000
## 44 AR-176 3 0.0000000
## 45 AR-192 2 0.6883825
## 46 AR-214 2 0.0000000
## 47 AR-237 2 0.6339993
## 48 AR-274 2 0.0000000
## 49 AR-300 2 0.0000000
## 50 AR-345 2 0.0000000
## 51 AR-346 2 0.8462352
## 52 AR-348 2 0.7479272
## 53 AR-382 2 0.7451339
## Posterior_marginal_prob_L Posterior_marginal_prob_I
## 1 0.20029039 7.997096e-01
## 2 0.24236975 1.780460e-08
## 3 0.13769233 2.870620e-08
## 4 0.04708870 9.529113e-01
## 5 0.99999919 8.095861e-07
## 6 0.09761816 9.023818e-01
## 7 0.49131418 5.086858e-01
## 8 0.22782070 3.361901e-06
## 9 0.22782283 1.232510e-06
## 10 0.15668331 8.433167e-01
## 11 0.26708459 7.329154e-01
## 12 0.09302359 9.069764e-01
## 13 0.95411195 4.588805e-02
## 14 0.10514624 8.948538e-01
## 15 0.20861539 1.457631e-06
## 16 0.62091008 3.790899e-01
## 17 0.05713046 9.428695e-01
## 18 0.89692628 1.030737e-01
## 19 0.99712108 2.878920e-03
## 20 0.20073604 7.992640e-01
## 21 0.18443518 8.155648e-01
## 22 0.34465510 6.553449e-01
## 23 0.99999318 6.820487e-06
## 24 0.99999371 6.292255e-06
## 25 0.05497134 9.450287e-01
## 26 0.25142969 2.054618e-08
## 27 0.25207193 8.834381e-07
## 28 0.13805319 4.061912e-04
## 29 0.99995771 4.228845e-05
## 30 0.31585812 6.841419e-01
## 31 0.99847214 1.527856e-03
## 32 0.09188394 9.081161e-01
## 33 0.87083346 1.291665e-01
## 34 0.99922528 7.747186e-04
## 35 0.85959432 1.404057e-01
## 36 0.22064240 7.793576e-01
## 37 0.31818593 6.818141e-01
## 38 0.99994032 5.967860e-05
## 39 0.26710484 7.328952e-01
## 40 0.31219012 1.300991e-06
## 41 0.06281198 9.371880e-01
## 42 0.99999479 5.206603e-06
## 43 0.25588975 7.441102e-01
## 44 0.99999554 4.459706e-06
## 45 0.31161734 1.483804e-07
## 46 0.71125930 2.887407e-01
## 47 0.36517961 8.210690e-04
## 48 0.09302472 9.069753e-01
## 49 0.09302334 9.069767e-01
## 50 0.09507480 9.049252e-01
## 51 0.15376481 2.089266e-11
## 52 0.25207193 8.834381e-07
## 53 0.25485692 9.180472e-06The joint probabilites give us values for each possible combination of states, depending on the number of recurrent episodes experienced by the participant.
joint_summary <- indiv_posteriors_joint %>%
group_by(subject_id, state_pair, joint_probability) %>%
filter(episode_number == max(episode_number)) %>%
mutate(percentage = round(joint_probability*100, 3),
total_recurrences = episode_number-1) %>%
select(-episode_number) %>%
arrange(subject_id, total_recurrences, percentage) %>%
relocate(total_recurrences, .before = state_pair)
joint_summary## # A tibble: 282 × 5
## # Groups: subject_id, state_pair, joint_probability [282]
## subject_id total_recurrences state_pair joint_probability percentage
## <chr> <dbl> <chr> <dbl> <dbl>
## 1 AR-024 4 CCCC 0 0
## 2 AR-024 4 LCCC 0 0
## 3 AR-024 4 ICCC 0 0
## 4 AR-024 4 CLCC 0 0
## 5 AR-024 4 LLCC 0 0
## 6 AR-024 4 ILCC 0 0
## 7 AR-024 4 CICC 0 0
## 8 AR-024 4 LICC 0 0
## 9 AR-024 4 IICC 0 0
## 10 AR-024 4 CCLC 0 0
## # ℹ 272 more rowsPlot marginal probabilities
marginal_summary <- indiv_posteriors_marginal %>%
pivot_longer(cols = !subject_id & !episode_number,
names_to = "posterior_type",
values_to = "posterior_value") %>%
mutate(posterior_classification = case_when(posterior_type == "Posterior_marginal_prob_C" ~ "Recrudescence",
posterior_type == "Posterior_marginal_prob_L" ~ "Relapse",
posterior_type == "Posterior_marginal_prob_I" ~ "Reinfection"),
posterior_classification = factor(posterior_classification,
levels = c("Relapse", "Recrudescence", "Reinfection"))) %>%
select(-posterior_type)
marginal_summary## # A tibble: 159 × 4
## subject_id episode_number posterior_value posterior_classification
## <chr> <int> <dbl> <fct>
## 1 AR-024 2 0 Recrudescence
## 2 AR-024 2 0.200 Relapse
## 3 AR-024 2 0.800 Reinfection
## 4 AR-024 3 0.758 Recrudescence
## 5 AR-024 3 0.242 Relapse
## 6 AR-024 3 0.0000000178 Reinfection
## 7 AR-024 4 0.862 Recrudescence
## 8 AR-024 4 0.138 Relapse
## 9 AR-024 4 0.0000000287 Reinfection
## 10 AR-024 5 0 Recrudescence
## # ℹ 149 more rowsBy episode number
marginal_summary %>%
group_by(subject_id) %>%
arrange(desc(posterior_value), .by_group = TRUE) %>%
mutate(subject_id = factor(subject_id, levels = unique(subject_id[order(posterior_value, decreasing = TRUE)]))) %>%
ggplot(aes(x = factor(episode_number), y = posterior_value, group = episode_number, fill = posterior_classification)) +
geom_bar(stat = "identity", position = "fill") +
scale_fill_manual(values = c("Relapse" = "turquoise3",
"Recrudescence" = "skyblue4",
"Reinfection" = "magenta3")) +
scale_x_discrete(expand = c(0, 0)) +
scale_y_continuous(expand = c(0, 0)) +
labs(x = "Episode number",
y = "Posterior probability",
fill = "") +
theme_bw() +
facet_wrap(~subject_id)
By treatment arm
marginal_summary %>%
left_join(analysis_data %>% select(subject_id, treatment_arm) %>% distinct(),
by = "subject_id") %>%
group_by(subject_id) %>%
arrange(desc(posterior_value), .by_group = TRUE) %>%
mutate(subject_id = factor(subject_id, levels = unique(subject_id[order(posterior_value, decreasing = TRUE)]))) %>%
ggplot(aes(x = factor(episode_number), y = posterior_value, group = episode_number, fill = posterior_classification)) +
geom_bar(stat = "identity", position = "fill") +
scale_fill_manual(values = c("Relapse" = "turquoise3",
"Recrudescence" = "skyblue4",
"Reinfection" = "magenta3")) +
scale_x_discrete(expand = c(0, 0)) +
scale_y_continuous(expand = c(0, 0)) +
labs(x = "Episode number",
y = "Posterior probability",
fill = "") +
theme_bw() +
facet_wrap(treatment_arm ~ subject_id)
By days since last episode
marginal_summary %>%
left_join(analysis_data %>% select(subject_id, episode_number, treatment_arm, visit_date, days_since_enrolment, days_since_last_episode) %>% distinct(),
by = c("subject_id", "episode_number")) %>%
group_by(subject_id) %>%
arrange(desc(posterior_value), .by_group = TRUE) %>%
mutate(subject_id = factor(subject_id, levels = unique(subject_id[order(posterior_value, decreasing = TRUE)]))) %>%
ggplot(aes(x = days_since_enrolment, y = posterior_value, group = episode_number, fill = posterior_classification)) +
geom_bar(stat = "identity", position = "fill", width = 3) +
scale_fill_manual(values = c("Relapse" = "turquoise3",
"Recrudescence" = "skyblue4",
"Reinfection" = "magenta3")) +
scale_x_continuous(expand = c(0, 0)) +
scale_y_continuous(expand = c(0, 0)) +
labs(x = "Days since enrolment",
y = "Posterior probability",
fill = "") +
theme_bw() +
facet_wrap(~subject_id)## Warning: `position_stack()` requires non-overlapping x intervals.
Plot joint probabilities
pal1 <- c(
"C" = "turquoise3",
"I" = "magenta3",
"L" = "skyblue4")
pal2 <- c(
"II" = "magenta3",
"IL" = "darkorange2",
"LI" = "goldenrod2",
"LL" = "skyblue4",
"CI" = "saddlebrown",
"CL" = "turquoise4",
"IC" = "darkolivegreen",
"LC" = "lightseagreen"
)
pal3 <- c(
"IIL" = "darkorange3",
"ILL" = "orangered3",
"LIL" = "goldenrod3",
"LLL" = "skyblue4",
"CCI" = "sienna4",
"CCL" = "cadetblue3",
"CLI" = "darkcyan",
"CLL" = "steelblue3",
"LCI" = "peru",
"LCL" = "powderblue",
"LLI" = "orange3"
)
pal4 <- c(
"ICCI" = "saddlebrown",
"ICCL" = "darkseagreen",
"ICLI" = "orange3",
"ICLL" = "lightsteelblue",
"ILCI" = "darkkhaki",
"ILCL" = "skyblue3",
"ILLI" = "gold3",
"LCCI" = "rosybrown",
"LCCL" = "lightblue3",
"LCLI" = "goldenrod3",
"LCLL" = "deepskyblue3",
"LLCI" = "tan",
"LLCL" = "mediumturquoise",
"LLLI" = "darkorange3"
)
combined_pal <- c(pal1, pal2, pal3, pal4)plotJointProb <- function(data, recurrence_n, prob_threshold, color_palette, ...){
data %>%
filter(total_recurrences == recurrence_n, joint_probability > prob_threshold) %>%
group_by(subject_id) %>%
mutate(state_pair = fct_reorder(state_pair, joint_probability)) %>%
ungroup() %>%
ggplot(aes(x = joint_probability,
y = subject_id,
fill = state_pair)) +
geom_bar(position = "stack", stat = "identity") +
scale_x_continuous(expand = c(0, 0)) +
scale_y_discrete(expand = c(0, 0)) +
scale_fill_manual(values = color_palette) +
labs(x = "Joint probability estimate",
y = "Participant",
fill = "Classification state") +
theme_bw() +
facet_wrap(~total_recurrences, scales = "free_y")
}By total number of recurrences
plot_joint1 <- plotJointProb(joint_summary, 1, 0, pal1)
plot_joint2 <- plotJointProb(joint_summary, 2, 0, pal2) + guides(fill = guide_legend(ncol = 2))
plot_joint3 <- plotJointProb(joint_summary, 3, 0.001, pal3) + guides(fill = guide_legend(ncol = 3))
plot_joint4 <- plotJointProb(joint_summary, 4, 0.001, pal4) + guides(fill = guide_legend(ncol = 3))
(plot_joint1 | plot_joint2) / (plot_joint3 | plot_joint4) +
plot_annotation(subtitle = "I = Reinfection, L = Relapse, C = Recrudescence",
caption = expression(italic("For >2 recurrences, only probability estimates > 0.001 are shown"))) +
plot_layout(heights = c(2, 1))
By treatment arm and number of recurrences
joint_summary %>%
left_join(analysis_data %>% select(subject_id, treatment_arm) %>% distinct(),
by = "subject_id") %>%
filter(joint_probability > 0.001) %>%
group_by(subject_id) %>%
mutate(state_pair = fct_reorder(state_pair, joint_probability)) %>%
ungroup() %>%
ggplot(aes(x = joint_probability,
y = reorder(subject_id, total_recurrences),
fill = state_pair)) +
geom_bar(position = "stack", stat = "identity") +
scale_x_continuous(expand = c(0, 0)) +
scale_y_discrete(expand = c(0, 0)) +
scale_fill_manual(values = combined_pal) +
labs(x = "Joint probability estimate",
y = "Participant",
fill = "Classification state",
subtitle = "I = Reinfection, L = Relapse, C = Recrudescence",
caption = expression(italic("Color coding: if > 1 recurrence, blue shades denote combinations of relapse and recrudescence (L and C), shades of orange/browns denote \ncombos of reinfection and relapse (I and L) and green shades denote combos of reinfection with recrudescence (I and C)"))) +
theme_bw() +
facet_wrap(treatment_arm~total_recurrences, scales = "free") +
# facet_wrap(total_recurrences~treatment_arm, scales = "free") +
theme(plot.caption = element_text(hjust = 0, vjust = -2)) 
‘Confidence’ in highest joint probability estimate
What is the joint probability estimate for the ‘highest ranking’ state pair classification? I.e. if >80% relatively high confidence? n=19
joint_summary %>%
filter(joint_probability > 0.001) %>%
group_by(subject_id) %>%
# get the highest prob for each subj and store ranking
mutate(ranking = rank(-joint_probability, ties.method = "first")) %>%
arrange(subject_id, ranking) %>%
filter(ranking == 1) %>%
ggplot(aes(x = reorder(subject_id, -joint_probability),
y = joint_probability,
fill = factor(total_recurrences))) +
geom_col() +
geom_hline(yintercept = 0.8, color = "darkgray", linetype = "dashed") +
labs(x = "Participant",
y = "Joint probability estimate for highest 'ranking' state pair",
fill = "Number of recurrent episodes",
caption = expression(italic("Dashed line indicates an arbitrary threshold of 80%"))) +
theme_bw() +
theme(axis.text.x = element_text(angle = 90, hjust = 1, vjust = 0.5))
How many state pairs make up the arbitrary 80% threshold?
joint_summary %>%
filter(joint_probability > 0.001) %>%
group_by(subject_id) %>%
# Get the highest prob for each subject and store ranking
mutate(ranking = rank(-joint_probability, ties.method = "first")) %>%
# reverse ranking order of ranking so we plot the highest rank at the bottom
arrange(subject_id, desc(ranking)) %>%
mutate(ranking = factor(ranking, levels = unique(ranking))) %>%
ggplot(aes(x = subject_id,
y = joint_probability,
fill = ranking)) +
geom_bar(stat = "identity", position = "stack") +
geom_hline(yintercept = 0.8, color = "darkgray", linetype = "dashed") +
scale_fill_manual(values = colorRampPalette(brewer.pal(12, "Paired"))(14),
guide = guide_legend(reverse = T)) +
scale_x_discrete(expand = c(0, 0)) +
scale_y_continuous(expand = c(0, 0)) +
labs(x = "Participant",
y = "Joint probability estimate for state pair",
fill = "Number of possible \nestimated state pairs",
caption = expression(italic("Dashed line indicates an arbitrary threshold of 80%"))) +
theme_bw() +
theme(axis.text.x = element_text(angle = 90, hjust = 1, vjust = 0.5)) +
facet_wrap(~total_recurrences, scales = "free_x")
Highest ranking classification by episode number
joint_top3_summary <- joint_summary %>%
select(-percentage) %>%
group_by(subject_id) %>%
# Get the highest prob for each subject and store ranking
mutate(ranking = rank(-joint_probability, ties.method = "first")) %>%
arrange(subject_id, ranking) %>%
filter(ranking %in% c(1, 2, 3))We will keep the ‘highest ranking’ joint probability estimate so we can compare the state classification for each episode with other metrics
joint_recurrence_summary <- joint_top3_summary %>%
select(-total_recurrences) %>%
filter(ranking == 1) %>%
mutate(state_pair_split = strsplit(state_pair, "")) %>% # Split state_pair into individual letters
unnest(state_pair_split) %>% # Unnest to create a new row for each letter
group_by(subject_id) %>%
mutate(
episode_number = row_number() + 1, # Create episode number starting from n+1 to reflect number of recurrence
joint_probability = joint_probability, # Keep the same joint_probability for all episodes\
state_classification = case_when(state_pair_split == "L" ~ "Relapse",
state_pair_split == "C" ~ "Recrudescence",
state_pair_split == "I" ~ "Reinfection")) %>%
select(subject_id, episode_number, state_classification, joint_probability)joint_recurrence_summary %>%
ggplot(aes(x = factor(episode_number), y = joint_probability, fill = state_classification)) +
geom_col() +
scale_fill_manual(values = c("Relapse" = "turquoise3",
"Recrudescence" = "skyblue4",
"Reinfection" = "magenta3")) +
scale_x_discrete(expand = c(0, 0)) +
scale_y_continuous(expand = c(0, 0)) +
labs(x = "Episode number",
y = "Posterior joint probability",
fill = "") +
theme_bw() +
facet_wrap(~subject_id)
Comparison marginal vs joint
posterior_summary <- marginal_summary %>%
rename("marginal_posterior_probability" = "posterior_value",
"marginal_classification" = "posterior_classification") %>%
left_join(joint_recurrence_summary %>%
rename("joint_classification" = "state_classification",
"joint_posterior_probability" = "joint_probability"),
by = c("subject_id", "episode_number")) %>%
pivot_longer(cols = c(marginal_posterior_probability,
joint_posterior_probability,
marginal_classification,
joint_classification),
names_to = c("estimate_type", ".value"),
names_pattern = "(marginal|joint)_(posterior_probability|classification)") %>%
distinct() posterior_summary %>%
group_by(subject_id, episode_number) %>%
arrange(desc(posterior_probability), .by_group = TRUE) %>%
mutate(classification = factor(classification, levels = unique(classification))) %>%
ungroup() %>%
ggplot(aes(x = factor(episode_number), y = posterior_probability, group = episode_number)) +
geom_bar(data = . %>% filter(estimate_type == "marginal"),
aes(fill = classification),
stat = "identity",
position = "fill") +
geom_segment(data = . %>% filter(estimate_type == "joint"),
aes(x = as.numeric(factor(episode_number)) - 0.4,
xend = as.numeric(factor(episode_number)) + 0.4,
y = posterior_probability,
yend = posterior_probability,
color = classification),
size = 1.5) +
geom_point(data = . %>% filter(estimate_type == "joint"),
aes(color = classification,
fill = classification),
shape = 21) +
scale_fill_manual(values = c("Relapse" = "turquoise3",
"Recrudescence" = "skyblue4",
"Reinfection" = "magenta3")) +
scale_color_manual(values = c("Relapse" = "#006064",
"Recrudescence" = "#3B4F75",
"Reinfection" = "#8B008B")) +
scale_x_discrete(expand = c(0, 0)) +
scale_y_continuous(expand = c(0, 0)) +
labs(x = "Episode number",
y = "Posterior probability",
fill = "Marginal classification",
color = "Joint probability estimate \n and classification") +
theme_bw() +
facet_wrap(~subject_id)## Warning: Using `size` aesthetic for lines was deprecated in ggplot2 3.4.0.
## ℹ Please use `linewidth` instead.
## This warning is displayed once every 8 hours.
## Call `lifecycle::last_lifecycle_warnings()` to see where this warning was
## generated.
Comparison to IBS and IBD estimates
sols_pnh <- read.csv(here("data/final", "sols_PNH_data.csv"))
ibd_r_data <- readRDS(here("data/final", "sols_all_meta.rds"))
ibd_rhat_data <- readRDS(here("data/final", "sols_sig_meta.rds"))For direct comparison with the we can only look at the paired comparisons between baseline samples and follow-up samples, but not other pw comparisons for a given participant.
sols_ibd <- ibd_r_data %>%
# filter to include only paired samples, and only pw comparisons to baseline
filter(comparison_type == "paired", day_id2 == 0) %>%
mutate(subject_id = paste0("AR-", patientid1),
day_id1 = as.integer(day_id1),
day_id2 = as.integer(day_id2)) %>%
group_by(subject_id) %>%
arrange(day_id1) %>%
# create episode_number accounting for baseline as episode 1
mutate(episode_number = row_number() + 1) %>%
ungroup()classification_summary <- indiv_posteriors_marginal %>%
rename("posterior_probability_recrudescence" = "Posterior_marginal_prob_C",
"posterior_probability_relapse" = "Posterior_marginal_prob_L",
"posterior_probability_reinfection" = "Posterior_marginal_prob_I") %>%
# merge with epi variables
left_join(analysis_data %>% select(subject_id, sample_id, episode_number, treatment_arm, visit_date, days_since_enrolment, days_since_last_episode) %>% distinct(),
by = c("subject_id", "episode_number")) %>%
# merge with PNH data
left_join(sols_pnh, by = c("subject_id" = "Patient", "episode_number" = "recurrence")) %>%
rename("prop_same_haps_1minusPNH" = "Prop_het",
"pnh_range" = "identity",
"classification" = "Classification") %>%
# create bins for 1-PNH (IBS metric)
mutate(ibs_range = cut(prop_same_haps_1minusPNH,
breaks = c(0, 0.25, 0.5, 0.75, 1),
include.lowest = TRUE,
labels = c("0-0.25", "0.25-0.5", "0.5-0.75", "0.75-1"))) %>%
# merge with IBD r
left_join(sols_ibd,
by = c("subject_id", "episode_number")) %>%
# create bins for IBD r
mutate(ibd_range = cut(estimate,
breaks = c(0, 0.25, 0.5, 0.75, 1),
include.lowest = TRUE,
labels = c("0-0.25", "0.25-0.5", "0.5-0.75", "0.75-1"))) %>%
# merge with joint probabilities
left_join(joint_recurrence_summary %>%
rename("joint_classification" = "state_classification"),
by = c("subject_id", "episode_number"))
classification_summary## subject_id episode_number posterior_probability_recrudescence
## 1 AR-024 2 0.0000000
## 2 AR-024 3 0.7576302
## 3 AR-024 4 0.8623076
## 4 AR-024 5 0.0000000
## 5 AR-221 2 0.0000000
## 6 AR-343 2 0.0000000
## 7 AR-366 2 0.0000000
## 8 AR-187 2 0.7721759
## 9 AR-187 3 0.7721759
## 10 AR-187 4 0.0000000
## 11 AR-383 2 0.0000000
## 12 AR-288 2 0.0000000
## 13 AR-234 2 0.0000000
## 14 AR-306 2 0.0000000
## 15 AR-306 3 0.7913832
## 16 AR-302 2 0.0000000
## 17 AR-341 2 0.0000000
## 18 AR-326 2 0.0000000
## 19 AR-067 2 0.0000000
## 20 AR-322 2 0.0000000
## 21 AR-355 2 0.0000000
## 22 AR-081 2 0.0000000
## 23 AR-081 3 0.0000000
## 24 AR-248 2 0.0000000
## 25 AR-304 2 0.0000000
## 26 AR-160 2 0.7485703
## 27 AR-110 2 0.7479272
## 28 AR-133 2 0.8615406
## 29 AR-133 3 0.0000000
## 30 AR-140 2 0.0000000
## 31 AR-140 3 0.0000000
## 32 AR-142 2 0.0000000
## 33 AR-142 3 0.0000000
## 34 AR-142 4 0.0000000
## 35 AR-173 2 0.0000000
## 36 AR-328 2 0.0000000
## 37 AR-328 3 0.0000000
## 38 AR-325 2 0.0000000
## 39 AR-246 2 0.0000000
## 40 AR-250 2 0.6878086
## 41 AR-258 2 0.0000000
## 42 AR-098 2 0.0000000
## 43 AR-176 2 0.0000000
## 44 AR-176 3 0.0000000
## 45 AR-192 2 0.6883825
## 46 AR-214 2 0.0000000
## 47 AR-237 2 0.6339993
## 48 AR-274 2 0.0000000
## 49 AR-300 2 0.0000000
## 50 AR-345 2 0.0000000
## 51 AR-346 2 0.8462352
## 52 AR-348 2 0.7479272
## 53 AR-382 2 0.7451339
## posterior_probability_relapse posterior_probability_reinfection sample_id
## 1 0.20029039 7.997096e-01 AR-024-56
## 2 0.24236975 1.780460e-08 AR-024-70
## 3 0.13769233 2.870620e-08 AR-024-84
## 4 0.04708870 9.529113e-01 AR-024-112
## 5 0.99999919 8.095861e-07 AR-221-70
## 6 0.09761816 9.023818e-01 AR-343-112
## 7 0.49131418 5.086858e-01 AR-366-168
## 8 0.22782070 3.361901e-06 AR-187-42
## 9 0.22782283 1.232510e-06 AR-187-84
## 10 0.15668331 8.433167e-01 AR-187-168
## 11 0.26708459 7.329154e-01 AR-383-168
## 12 0.09302359 9.069764e-01 AR-288-70
## 13 0.95411195 4.588805e-02 AR-234-112
## 14 0.10514624 8.948538e-01 AR-306-84
## 15 0.20861539 1.457631e-06 AR-306-85
## 16 0.62091008 3.790899e-01 AR-302-70
## 17 0.05713046 9.428695e-01 AR-341-84
## 18 0.89692628 1.030737e-01 AR-326-168
## 19 0.99712108 2.878920e-03 AR-067-42
## 20 0.20073604 7.992640e-01 AR-322-140
## 21 0.18443518 8.155648e-01 AR-355-140
## 22 0.34465510 6.553449e-01 AR-081-140
## 23 0.99999318 6.820487e-06 AR-081-168
## 24 0.99999371 6.292255e-06 AR-248-84
## 25 0.05497134 9.450287e-01 AR-304-28
## 26 0.25142969 2.054618e-08 AR-160-84
## 27 0.25207193 8.834381e-07 AR-110-84
## 28 0.13805319 4.061912e-04 AR-133-84
## 29 0.99995771 4.228845e-05 AR-133-140
## 30 0.31585812 6.841419e-01 AR-140-56
## 31 0.99847214 1.527856e-03 AR-140-70
## 32 0.09188394 9.081161e-01 AR-142-42
## 33 0.87083346 1.291665e-01 AR-142-46
## 34 0.99922528 7.747186e-04 AR-142-97
## 35 0.85959432 1.404057e-01 AR-173-28
## 36 0.22064240 7.793576e-01 AR-328-84
## 37 0.31818593 6.818141e-01 AR-328-112
## 38 0.99994032 5.967860e-05 AR-325-140
## 39 0.26710484 7.328952e-01 AR-246-70
## 40 0.31219012 1.300991e-06 AR-250-56
## 41 0.06281198 9.371880e-01 AR-258-140
## 42 0.99999479 5.206603e-06 AR-098-56
## 43 0.25588975 7.441102e-01 AR-176-28
## 44 0.99999554 4.459706e-06 AR-176-33
## 45 0.31161734 1.483804e-07 AR-192-70
## 46 0.71125930 2.887407e-01 AR-214-64
## 47 0.36517961 8.210690e-04 AR-237-70
## 48 0.09302472 9.069753e-01 AR-274-56
## 49 0.09302334 9.069767e-01 AR-300-84
## 50 0.09507480 9.049252e-01 AR-345-112
## 51 0.15376481 2.089266e-11 AR-346-112
## 52 0.25207193 8.834381e-07 AR-348-84
## 53 0.25485692 9.180472e-06 AR-382-168
## treatment_arm visit_date days_since_enrolment days_since_last_episode
## 1 AL-PQ 2017-12-11 56 56
## 2 AL-PQ 2017-12-25 70 14
## 3 AL-PQ 2018-01-09 85 15
## 4 AL-PQ 2018-02-05 112 27
## 5 AL-PQ 2018-11-15 70 70
## 6 DHP-PQ 2019-05-08 112 112
## 7 DHP-PQ 2019-07-24 168 168
## 8 AL 2018-09-27 43 43
## 9 AL 2018-11-07 84 41
## 10 AL 2019-01-30 168 84
## 11 DHP-PQ 2019-08-07 168 168
## 12 DHP-PQ 2019-01-08 70 70
## 13 DHP-PQ 2019-01-04 112 112
## 14 AL-PQ 2019-02-18 84 84
## 15 AL-PQ 2019-02-19 85 1
## 16 DHP-PQ 2019-01-23 70 70
## 17 DHP-PQ 2019-04-08 84 84
## 18 DHP-PQ 2019-06-26 168 168
## 19 AL 2018-01-22 42 42
## 20 DHP-PQ 2019-05-28 140 140
## 21 AL-PQ 2019-06-13 140 140
## 22 AL-PQ 2018-07-16 140 140
## 23 AL-PQ 2018-08-13 168 28
## 24 AL-PQ 2018-12-22 82 82
## 25 AL 2018-12-17 28 28
## 26 AL 2018-10-16 84 84
## 27 AL-PQ 2018-08-07 84 84
## 28 AL 2018-09-10 84 84
## 29 AL 2018-11-05 140 56
## 30 AL-PQ 2018-08-20 56 56
## 31 AL-PQ 2018-09-03 70 14
## 32 AL 2018-08-09 42 42
## 33 AL 2018-08-13 46 4
## 34 AL 2018-10-03 97 51
## 35 AL 2018-09-03 28 28
## 36 DHP-PQ 2019-04-01 82 82
## 37 DHP-PQ 2019-05-02 113 31
## 38 AL 2019-05-29 140 140
## 39 DHP-PQ 2018-12-05 70 70
## 40 AL 2018-11-26 56 56
## 41 DHP-PQ 2019-02-25 140 140
## 42 AL-PQ 2018-06-14 56 56
## 43 AL 2018-09-05 28 28
## 44 AL 2018-09-10 33 5
## 45 AL-PQ 2018-10-29 70 70
## 46 AL-PQ 2018-11-07 64 64
## 47 AL-PQ 2018-11-26 70 70
## 48 AL-PQ 2018-12-12 56 56
## 49 AL-PQ 2019-02-04 84 84
## 50 AL-PQ 2019-05-08 112 112
## 51 DHP-PQ 2019-05-08 112 112
## 52 AL 2019-04-11 84 84
## 53 AL-PQ 2019-08-06 168 168
## sample date treatment treatment2 prop_same_haps_1minusPNH
## 1 AR-024-56 2017-12-11 AL-PQ PQ 0.3068233
## 2 AR-024-70 2017-12-25 AL-PQ PQ 0.2219938
## 3 AR-024-84 2018-01-09 AL-PQ PQ 0.2414539
## 4 AR-024-112 2018-02-05 AL-PQ PQ 0.1420881
## 5 AR-221-70 2018-11-15 AL-PQ PQ 1.0000000
## 6 AR-343-112 2019-05-08 DHP-PQ PQ 0.4906786
## 7 AR-366-168 2019-07-24 DHP-PQ PQ 0.4117348
## 8 AR-187-42 2018-09-27 AL AL 1.0000000
## 9 AR-187-84 2018-11-07 AL AL 1.0000000
## 10 AR-187-168 2019-01-30 AL AL 0.3852387
## 11 AR-383-168 2019-08-07 DHP-PQ PQ 0.2490546
## 12 AR-288-70 2019-01-08 DHP-PQ PQ 0.3338144
## 13 AR-234-112 2019-01-04 DHP-PQ PQ 0.8279783
## 14 AR-306-84 2019-02-18 AL-PQ PQ 0.7195963
## 15 AR-306-85 2019-02-19 AL-PQ PQ 0.7195963
## 16 AR-302-70 2019-01-23 DHP-PQ PQ 0.8824741
## 17 AR-341-84 2019-04-08 DHP-PQ PQ 0.4394276
## 18 AR-326-168 2019-06-26 DHP-PQ PQ 0.8963978
## 19 AR-067-42 2018-01-22 AL AL 1.0000000
## 20 AR-322-140 2019-05-28 DHP-PQ PQ 0.3955032
## 21 AR-355-140 2019-06-13 AL-PQ PQ 0.3345985
## 22 AR-081-140 2018-07-16 AL-PQ PQ 0.6640732
## 23 AR-081-168 2018-08-13 AL-PQ PQ 0.6640732
## 24 AR-248-84 2018-12-22 AL-PQ PQ 1.0000000
## 25 AR-304-28 2018-12-17 AL AL 0.5245581
## 26 AR-160-84 2018-10-16 AL AL 1.0000000
## 27 AR-110-84 2018-08-07 AL-PQ PQ 1.0000000
## 28 AR-133-84 2018-09-10 AL AL 1.0000000
## 29 AR-133-140 2018-11-05 AL AL 1.0000000
## 30 AR-140-56 2018-08-20 AL-PQ PQ 0.8158586
## 31 AR-140-70 2018-09-03 AL-PQ PQ 1.0000000
## 32 AR-142-42 2018-08-09 AL AL 0.3976700
## 33 AR-142-46 2018-08-13 AL AL 0.4526802
## 34 AR-142-97 2018-10-03 AL AL 0.7944225
## 35 AR-173-28 2018-09-03 AL AL 0.8550821
## 36 AR-328-84 2019-04-01 DHP-PQ PQ 0.4645095
## 37 AR-328-112 2019-05-02 DHP-PQ PQ 0.6741595
## 38 AR-325-140 2019-05-29 AL AL 0.8956478
## 39 AR-246-70 2018-12-05 DHP-PQ PQ 0.2168451
## 40 AR-250-56 2018-11-26 AL AL 1.0000000
## 41 AR-258-140 2019-02-25 DHP-PQ PQ 0.8162594
## 42 AR-098-56 2018-06-14 AL-PQ PQ 1.0000000
## 43 AR-176-28 2018-09-05 AL AL 0.4092557
## 44 AR-176-33 2018-09-10 AL AL 0.3871925
## 45 AR-192-70 2018-10-29 AL-PQ PQ 1.0000000
## 46 AR-214-64 2018-11-07 AL-PQ PQ 0.9085737
## 47 AR-237-70 2018-11-26 AL-PQ PQ 1.0000000
## 48 AR-274-56 2018-12-12 AL-PQ PQ 0.1967818
## 49 AR-300-84 2019-02-04 AL-PQ PQ 0.2981501
## 50 AR-345-112 2019-05-08 AL-PQ PQ 0.3738918
## 51 AR-346-112 2019-05-08 DHP-PQ PQ 1.0000000
## 52 AR-348-84 2019-04-11 AL AL 1.0000000
## 53 AR-382-168 2019-08-06 AL-PQ PQ 1.0000000
## delay_since_prev_ep change_moi max_moi clones pnh_range classification
## 1 56 1 2 poly (0,0.75] Homologous
## 2 70 -1 1 mono (0.75,1] Heterologous
## 3 85 0 1 mono (0.75,1] Heterologous
## 4 112 1 2 poly (0.75,1] Heterologous
## 5 70 1 2 poly (-Inf,0] Homologous
## 6 112 1 3 poly (0,0.75] Homologous
## 7 168 -2 1 mono (0,0.75] Homologous
## 8 43 0 1 mono (-Inf,0] Homologous
## 9 84 0 1 mono (-Inf,0] Homologous
## 10 168 1 2 poly (0,0.75] Homologous
## 11 168 0 1 mono (0.75,1] Heterologous
## 12 70 0 2 poly (0,0.75] Homologous
## 13 112 -1 2 poly (0,0.75] Homologous
## 14 84 0 2 poly (0,0.75] Homologous
## 15 85 -1 1 mono (0,0.75] Homologous
## 16 70 0 2 poly (0,0.75] Homologous
## 17 84 1 3 poly (0,0.75] Homologous
## 18 168 -1 2 poly (0,0.75] Homologous
## 19 42 0 2 poly (-Inf,0] Homologous
## 20 140 1 2 poly (0,0.75] Homologous
## 21 140 1 2 poly (0,0.75] Homologous
## 22 140 0 2 poly (0,0.75] Homologous
## 23 168 1 3 poly (0,0.75] Homologous
## 24 82 0 2 poly (-Inf,0] Homologous
## 25 28 -1 2 poly (0,0.75] Homologous
## 26 84 0 1 mono (-Inf,0] Homologous
## 27 84 0 1 mono (-Inf,0] Homologous
## 28 84 0 1 mono (-Inf,0] Homologous
## 29 140 2 3 poly (-Inf,0] Homologous
## 30 56 1 3 poly (0,0.75] Homologous
## 31 70 -1 2 poly (-Inf,0] Homologous
## 32 42 1 3 poly (0,0.75] Homologous
## 33 46 -2 1 mono (0,0.75] Homologous
## 34 97 1 2 poly (0,0.75] Homologous
## 35 28 -1 1 mono (0,0.75] Homologous
## 36 82 -1 1 mono (0,0.75] Homologous
## 37 113 1 2 poly (0,0.75] Homologous
## 38 140 0 1 mono (0,0.75] Homologous
## 39 70 0 1 mono (0.75,1] Heterologous
## 40 56 -1 1 mono (-Inf,0] Homologous
## 41 140 -1 2 poly (0,0.75] Homologous
## 42 56 1 2 poly (-Inf,0] Homologous
## 43 28 1 2 poly (0,0.75] Homologous
## 44 33 0 2 poly (0,0.75] Homologous
## 45 70 -1 1 mono (-Inf,0] Homologous
## 46 64 -2 2 poly (0,0.75] Homologous
## 47 70 -1 1 mono (-Inf,0] Homologous
## 48 56 0 2 poly (0.75,1] Heterologous
## 49 84 0 2 poly (0,0.75] Homologous
## 50 112 0 2 poly (0,0.75] Homologous
## 51 112 -1 2 poly (-Inf,0] Homologous
## 52 84 0 1 mono (-Inf,0] Homologous
## 53 168 0 1 mono (-Inf,0] Homologous
## .group ibs_range sampleid1 sampleid2 estimate p_value CI_lower
## 1 1 0.25-0.5 AR-024-56 AR-024-0 0.000 5.000000e-01 0.000
## 2 1 0-0.25 AR-024-70 AR-024-0 0.000 5.000000e-01 0.000
## 3 1 0-0.25 AR-024-84 AR-024-0 0.000 5.000000e-01 0.000
## 4 1 0-0.25 AR-024-112 AR-024-0 0.000 5.000000e-01 0.000
## 5 15 0.75-1 AR-221-70 AR-221-0 1.000 6.314730e-08 0.732
## 6 33 0.25-0.5 AR-343-112 AR-343-0 0.000 5.000000e-01 0.000
## 7 39 0.25-0.5 AR-366-168 AR-366-0 0.282 1.139043e-02 0.022
## 8 12 0.75-1 AR-187-42 AR-187-0 1.000 1.236722e-06 0.712
## 9 12 0.75-1 AR-187-84 AR-187-0 1.000 7.824760e-08 0.763
## 10 12 0.25-0.5 AR-187-168 AR-187-0 0.175 2.099312e-01 0.000
## 11 41 0-0.25 AR-383-168 AR-383-0 0.000 5.000000e-01 0.000
## 12 23 0.25-0.5 AR-288-70 AR-288-0 0.000 5.000000e-01 0.000
## 13 16 0.75-1 AR-234-112 AR-234-0 0.669 1.240223e-03 0.212
## 14 27 0.5-0.75 AR-306-84 AR-306-0 0.548 6.480622e-03 0.103
## 15 27 0.5-0.75 AR-306-85 AR-306-0 0.567 3.426794e-03 0.138
## 16 25 0.75-1 AR-302-70 AR-302-0 0.560 3.842267e-02 0.000
## 17 32 0.25-0.5 AR-341-84 AR-341-0 0.000 5.000000e-01 0.000
## 18 30 0.75-1 AR-326-168 AR-326-0 0.725 2.205554e-02 0.022
## 19 2 0.75-1 AR-067-42 AR-067-0 1.000 3.856190e-05 0.631
## 20 28 0.25-0.5 AR-322-140 AR-322-0 0.000 5.000000e-01 0.000
## 21 37 0.25-0.5 AR-355-140 AR-355-0 0.031 4.380358e-01 0.000
## 22 3 0.5-0.75 AR-081-140 AR-081-0 0.000 5.000000e-01 0.000
## 23 3 0.5-0.75 AR-081-168 AR-081-0 0.000 5.000000e-01 0.000
## 24 19 0.75-1 AR-248-84 AR-248-0 1.000 5.548439e-06 0.697
## 25 26 0.5-0.75 AR-304-28 AR-304-0 0.000 5.000000e-01 0.000
## 26 9 0.75-1 AR-160-84 AR-160-0 1.000 8.884332e-10 0.775
## 27 5 0.75-1 AR-110-84 AR-110-0 1.000 2.540399e-08 0.768
## 28 6 0.75-1 AR-133-84 AR-133-0 1.000 3.057878e-06 0.662
## 29 6 0.75-1 AR-133-140 AR-133-0 1.000 4.752026e-07 0.718
## 30 7 0.75-1 AR-140-56 AR-140-0 0.683 3.407867e-02 0.000
## 31 7 0.75-1 AR-140-70 AR-140-0 1.000 1.578781e-04 0.585
## 32 8 0.25-0.5 AR-142-42 AR-142-0 0.000 5.000000e-01 0.000
## 33 8 0.25-0.5 AR-142-46 AR-142-0 0.000 5.000000e-01 0.000
## 34 8 0.75-1 AR-142-97 AR-142-0 0.672 9.258504e-04 0.222
## 35 10 0.75-1 AR-173-28 AR-173-0 0.755 3.153185e-03 0.200
## 36 31 0.25-0.5 AR-328-84 AR-328-0 0.111 2.890700e-01 0.000
## 37 31 0.5-0.75 AR-328-112 AR-328-0 0.374 4.673184e-02 0.000
## 38 29 0.75-1 AR-325-140 AR-325-0 0.863 1.458441e-07 0.501
## 39 18 0-0.25 AR-246-70 AR-246-0 0.000 5.000000e-01 0.000
## 40 20 0.75-1 AR-250-56 AR-250-0 1.000 5.050400e-08 0.764
## 41 21 0.75-1 AR-258-140 AR-258-0 0.659 4.000648e-04 0.218
## 42 4 0.75-1 AR-098-56 AR-098-0 1.000 5.196368e-08 0.755
## 43 11 0.25-0.5 AR-176-28 AR-176-0 0.000 5.000000e-01 0.000
## 44 11 0.25-0.5 AR-176-33 AR-176-0 0.000 5.000000e-01 0.000
## 45 13 0.75-1 AR-192-70 AR-192-0 1.000 8.370809e-09 0.764
## 46 14 0.75-1 AR-214-64 AR-214-0 0.672 7.405563e-03 0.089
## 47 17 0.75-1 AR-237-70 AR-237-0 1.000 3.886609e-04 0.536
## 48 22 0-0.25 AR-274-56 AR-274-0 0.000 5.000000e-01 0.000
## 49 24 0.25-0.5 AR-300-84 AR-300-0 0.090 3.160523e-01 0.000
## 50 34 0.25-0.5 AR-345-112 AR-345-0 0.000 5.000000e-01 0.000
## 51 35 0.75-1 AR-346-112 AR-346-0 1.000 1.784710e-06 0.713
## 52 36 0.75-1 AR-348-84 AR-348-0 1.000 2.540399e-08 0.768
## 53 40 0.75-1 AR-382-168 AR-382-0 1.000 7.855351e-07 0.705
## CI_upper patientid1 patientid2 day_id1 day_id2 comparison_type
## 1 0.341 024 024 56 0 paired
## 2 0.300 024 024 70 0 paired
## 3 0.346 024 024 84 0 paired
## 4 0.251 024 024 112 0 paired
## 5 1.000 221 221 70 0 paired
## 6 0.442 343 343 112 0 paired
## 7 0.635 366 366 168 0 paired
## 8 1.000 187 187 42 0 paired
## 9 1.000 187 187 84 0 paired
## 10 0.656 187 187 168 0 paired
## 11 0.447 383 383 168 0 paired
## 12 0.388 288 288 70 0 paired
## 13 0.938 234 234 112 0 paired
## 14 0.872 306 306 84 0 paired
## 15 0.877 306 306 85 0 paired
## 16 0.963 302 302 70 0 paired
## 17 0.389 341 341 84 0 paired
## 18 0.983 326 326 168 0 paired
## 19 1.000 067 067 42 0 paired
## 20 0.404 322 322 140 0 paired
## 21 0.482 355 355 140 0 paired
## 22 0.570 081 081 140 0 paired
## 23 0.498 081 081 168 0 paired
## 24 1.000 248 248 84 0 paired
## 25 0.474 304 304 28 0 paired
## 26 1.000 160 160 84 0 paired
## 27 1.000 110 110 84 0 paired
## 28 1.000 133 133 84 0 paired
## 29 1.000 133 133 140 0 paired
## 30 0.980 140 140 56 0 paired
## 31 1.000 140 140 70 0 paired
## 32 0.494 142 142 42 0 paired
## 33 0.542 142 142 46 0 paired
## 34 0.939 142 142 97 0 paired
## 35 0.984 173 173 28 0 paired
## 36 0.575 328 328 84 0 paired
## 37 0.807 328 328 112 0 paired
## 38 0.991 325 325 140 0 paired
## 39 0.456 246 246 70 0 paired
## 40 1.000 250 250 56 0 paired
## 41 0.935 258 258 140 0 paired
## 42 1.000 098 098 56 0 paired
## 43 0.489 176 176 28 0 paired
## 44 0.411 176 176 33 0 paired
## 45 1.000 192 192 70 0 paired
## 46 0.978 214 214 64 0 paired
## 47 1.000 237 237 70 0 paired
## 48 0.300 274 274 56 0 paired
## 49 0.547 300 300 84 0 paired
## 50 0.480 345 345 112 0 paired
## 51 1.000 346 346 112 0 paired
## 52 1.000 348 348 84 0 paired
## 53 1.000 382 382 168 0 paired
## pair sig_est ibd_range joint_classification
## 1 AR-024-56 / AR-024-0 not significant 0-0.25 Reinfection
## 2 AR-024-70 / AR-024-0 not significant 0-0.25 Recrudescence
## 3 AR-024-84 / AR-024-0 not significant 0-0.25 Recrudescence
## 4 AR-024-112 / AR-024-0 not significant 0-0.25 Reinfection
## 5 AR-221-70 / AR-221-0 significant 0.75-1 Relapse
## 6 AR-343-112 / AR-343-0 not significant 0-0.25 Reinfection
## 7 AR-366-168 / AR-366-0 significant 0.25-0.5 Reinfection
## 8 AR-187-42 / AR-187-0 significant 0.75-1 Recrudescence
## 9 AR-187-84 / AR-187-0 significant 0.75-1 Recrudescence
## 10 AR-187-168 / AR-187-0 not significant 0-0.25 Reinfection
## 11 AR-383-168 / AR-383-0 not significant 0-0.25 Reinfection
## 12 AR-288-70 / AR-288-0 not significant 0-0.25 Reinfection
## 13 AR-234-112 / AR-234-0 significant 0.5-0.75 Relapse
## 14 AR-306-84 / AR-306-0 significant 0.5-0.75 Reinfection
## 15 AR-306-85 / AR-306-0 significant 0.5-0.75 Recrudescence
## 16 AR-302-70 / AR-302-0 significant 0.5-0.75 Relapse
## 17 AR-341-84 / AR-341-0 not significant 0-0.25 Reinfection
## 18 AR-326-168 / AR-326-0 significant 0.5-0.75 Relapse
## 19 AR-067-42 / AR-067-0 significant 0.75-1 Relapse
## 20 AR-322-140 / AR-322-0 not significant 0-0.25 Reinfection
## 21 AR-355-140 / AR-355-0 not significant 0-0.25 Reinfection
## 22 AR-081-140 / AR-081-0 not significant 0-0.25 Reinfection
## 23 AR-081-168 / AR-081-0 not significant 0-0.25 Relapse
## 24 AR-248-84 / AR-248-0 significant 0.75-1 Relapse
## 25 AR-304-28 / AR-304-0 not significant 0-0.25 Reinfection
## 26 AR-160-84 / AR-160-0 significant 0.75-1 Recrudescence
## 27 AR-110-84 / AR-110-0 significant 0.75-1 Recrudescence
## 28 AR-133-84 / AR-133-0 significant 0.75-1 Recrudescence
## 29 AR-133-140 / AR-133-0 significant 0.75-1 Relapse
## 30 AR-140-56 / AR-140-0 significant 0.5-0.75 Reinfection
## 31 AR-140-70 / AR-140-0 significant 0.75-1 Relapse
## 32 AR-142-42 / AR-142-0 not significant 0-0.25 Reinfection
## 33 AR-142-46 / AR-142-0 not significant 0-0.25 Relapse
## 34 AR-142-97 / AR-142-0 significant 0.5-0.75 Relapse
## 35 AR-173-28 / AR-173-0 significant 0.75-1 Relapse
## 36 AR-328-84 / AR-328-0 not significant 0-0.25 Reinfection
## 37 AR-328-112 / AR-328-0 significant 0.25-0.5 Reinfection
## 38 AR-325-140 / AR-325-0 significant 0.75-1 Relapse
## 39 AR-246-70 / AR-246-0 not significant 0-0.25 Reinfection
## 40 AR-250-56 / AR-250-0 significant 0.75-1 Recrudescence
## 41 AR-258-140 / AR-258-0 significant 0.5-0.75 Reinfection
## 42 AR-098-56 / AR-098-0 significant 0.75-1 Relapse
## 43 AR-176-28 / AR-176-0 not significant 0-0.25 Reinfection
## 44 AR-176-33 / AR-176-0 not significant 0-0.25 Relapse
## 45 AR-192-70 / AR-192-0 significant 0.75-1 Recrudescence
## 46 AR-214-64 / AR-214-0 significant 0.5-0.75 Relapse
## 47 AR-237-70 / AR-237-0 significant 0.75-1 Recrudescence
## 48 AR-274-56 / AR-274-0 not significant 0-0.25 Reinfection
## 49 AR-300-84 / AR-300-0 not significant 0-0.25 Reinfection
## 50 AR-345-112 / AR-345-0 not significant 0-0.25 Reinfection
## 51 AR-346-112 / AR-346-0 significant 0.75-1 Recrudescence
## 52 AR-348-84 / AR-348-0 significant 0.75-1 Recrudescence
## 53 AR-382-168 / AR-382-0 significant 0.75-1 Recrudescence
## joint_probability
## 1 0.5035808
## 2 0.5035808
## 3 0.5035808
## 4 0.5035808
## 5 0.9999992
## 6 0.9023818
## 7 0.5086858
## 8 0.4787060
## 9 0.4787060
## 10 0.4787060
## 11 0.7329154
## 12 0.9069764
## 13 0.9541120
## 14 0.7117713
## 15 0.7117713
## 16 0.6209101
## 17 0.9428695
## 18 0.8969263
## 19 0.9971211
## 20 0.7992640
## 21 0.8155648
## 22 0.6553398
## 23 0.6553398
## 24 0.9999937
## 25 0.9450287
## 26 0.7485703
## 27 0.7479272
## 28 0.8615093
## 29 0.8615093
## 30 0.6831029
## 31 0.6831029
## 32 0.7883076
## 33 0.7883076
## 34 0.7883076
## 35 0.8595943
## 36 0.5402922
## 37 0.5402922
## 38 0.9999403
## 39 0.7328952
## 40 0.6878086
## 41 0.9371880
## 42 0.9999948
## 43 0.7441066
## 44 0.7441066
## 45 0.6883825
## 46 0.7112593
## 47 0.6339993
## 48 0.9069753
## 49 0.9069767
## 50 0.9049252
## 51 0.8462352
## 52 0.7479272
## 53 0.7451339Comparison with IBS
classification_summary %>%
select(subject_id, sample_id, episode_number, visit_date, days_since_enrolment, days_since_last_episode, starts_with("posterior_"), starts_with("joint"), prop_same_haps_1minusPNH, pnh_range, ibs_range, classification) %>%
pivot_longer(cols = (starts_with("posterior")),
names_to = "posterior_type",
values_to = "posterior_value") %>%
mutate(posterior_range = cut(posterior_value,
breaks = c(0, 0.25, 0.5, 0.75, 1),
include.lowest = TRUE,
labels = c("0-0.25", "0.25-0.5", "0.5-0.75", "0.75-1"))) %>%
count(ibs_range, classification, posterior_type, posterior_range)Plot IBS range vs probabilistic classification (marginal probability)
classification_summary %>%
select(subject_id, sample_id, pair, episode_number, visit_date, days_since_enrolment, days_since_last_episode, starts_with("posterior_"), prop_same_haps_1minusPNH, pnh_range, ibs_range, classification) %>%
pivot_longer(cols = (starts_with("posterior")),
names_to = "posterior_type",
values_to = "posterior_value") %>%
mutate(posterior_range = cut(posterior_value,
breaks = c(0, 0.25, 0.5, 0.75, 1),
include.lowest = TRUE,
labels = c("0-0.25", "0.25-0.5", "0.5-0.75", "0.75-1"))) %>%
ggplot(aes(x = reorder(pair, episode_number), y = posterior_value, group = episode_number, fill = posterior_type)) +
geom_bar(stat = "identity", position = "fill") +
scale_fill_manual(values = c("turquoise3", "magenta3", "skyblue4"),
labels = c("posterior_probability_relapse" = "Relapse",
"posterior_probability_recrudescence" = "Recrudescence",
"posterior_probability_reinfection" = "Reinfection")) +
scale_x_discrete(expand = c(0, 0)) +
scale_y_continuous(expand = c(0, 0)) +
labs(x = "Sample pair",
y = "Posterior marginal probability",
fill = "") +
theme_bw() +
facet_wrap(~ibs_range, scales = "free_x") +
theme(axis.text.x = element_text(angle = 90, hjust = 1, vjust = 0.5))
### Plot IBS range vs probabilistic classification (joint probability,
highest ranking)
classification_summary %>%
select(subject_id, sample_id, pair, episode_number, visit_date, days_since_enrolment, days_since_last_episode, starts_with("joint"), prop_same_haps_1minusPNH, pnh_range, ibs_range, classification) %>%
ggplot(aes(x = reorder(pair, episode_number), y = joint_probability, group = episode_number, fill = joint_classification)) +
geom_bar(stat = "identity") +
scale_fill_manual(values = c("Relapse" = "turquoise3",
"Recrudescence" = "skyblue4",
"Reinfection" = "magenta3")) +
scale_x_discrete(expand = c(0, 0)) +
scale_y_continuous(expand = c(0, 0)) +
labs(x = "Sample pair",
y = "Posterior joint probability (highest ranking classification)",
fill = "") +
theme_bw() +
facet_wrap(~ibs_range, scales = "free_x") +
theme(axis.text.x = element_text(angle = 90, hjust = 1, vjust = 0.5))
Comparison with IBD
classification_summary %>%
select(subject_id, sample_id, episode_number, visit_date, days_since_enrolment, days_since_last_episode, starts_with("posterior_"), ibd_range, CI_lower, CI_upper, sig_est) %>%
pivot_longer(cols = (starts_with("posterior")),
names_to = "posterior_type",
values_to = "posterior_value") %>%
mutate(posterior_range = cut(posterior_value,
breaks = c(0, 0.25, 0.5, 0.75, 1),
include.lowest = TRUE,
labels = c("0-0.25", "0.25-0.5", "0.5-0.75", "0.75-1"))) %>%
count(ibd_range, sig_est, posterior_type, posterior_range)For reference - all pairwise comparisons
ibd_r_data %>%
filter(comparison_type == "paired") %>%
ggplot(aes(x = pair, y = estimate, group = patientid1, color = sig_est)) +
geom_point(size = 3) +
geom_errorbar(aes(ymin = CI_lower, ymax = CI_upper)) +
scale_color_manual(values = c("not significant" = "darkgrey",
"significant" = "indianred3")) +
facet_wrap(~patientid1, scales = "free_x") +
theme_bw() +
theme(axis.text.x = element_text(angle = 90, hjust = 1, vjust = 0.5))
Plot IBD estimate and CI for paired samples compared to baseline for each participant
classification_summary %>%
ggplot(aes(x = reorder(sampleid1, days_since_enrolment), y = estimate, group = subject_id, color = sig_est)) +
geom_point(size = 3) +
geom_errorbar(aes(ymin = CI_lower, ymax = CI_upper)) +
scale_color_manual(values = c("not significant" = "darkgrey",
"significant" = "indianred3")) +
labs(x = "Sample",
y = "IBD",
fill = "Significance") +
facet_wrap(~subject_id, scales = "free_x") +
theme_bw() +
theme(axis.text.x = element_text(angle = 90, hjust = 1, vjust = 0.5))
ibd_r_data %>%
filter(comparison_type == "not paired") %>%
ggplot(aes(x = pair, y = estimate, group = patientid1, color = sig_est)) +
geom_point(size = 3) +
geom_errorbar(aes(ymin = CI_lower, ymax = CI_upper)) +
scale_color_manual(values = c("not significant" = "darkgrey",
"significant" = "indianred3")) +
theme_bw() +
theme(axis.text.x = element_text(angle = 90, hjust = 1, vjust = 0.5))Plot IBD estimate and CI for paired samples compared to baseline for each participant, based on days since enrolment
classification_summary %>%
ggplot(aes(x = days_since_enrolment, y = estimate, group = subject_id, color = sig_est)) +
geom_point(size = 3) +
geom_errorbar(aes(ymin = CI_lower, ymax = CI_upper)) +
scale_color_manual(values = c("not significant" = "darkgrey",
"significant" = "indianred3")) +
labs(x = "Days since last episode",
y = "IBD",
fill = "Significance") +
facet_wrap(~subject_id, scales = "free_x") +
theme_bw() +
theme(axis.text.x = element_text(angle = 90, hjust = 1, vjust = 0.5))
Plot IBD range vs probabilistic classification (marginal)
classification_summary %>%
select(subject_id, sample_id, pair, episode_number, visit_date, days_since_enrolment, days_since_last_episode, starts_with("posterior_"), ibd_range, CI_lower, CI_upper, sig_est) %>%
pivot_longer(cols = (starts_with("posterior")),
names_to = "posterior_type",
values_to = "posterior_value") %>%
mutate(posterior_range = cut(posterior_value,
breaks = c(0, 0.25, 0.5, 0.75, 1),
include.lowest = TRUE,
labels = c("0-0.25", "0.25-0.5", "0.5-0.75", "0.75-1"))) %>%
ggplot(aes(x = reorder(pair, episode_number), y = posterior_value, group = episode_number, fill = posterior_type)) +
geom_bar(stat = "identity", position = "fill") +
scale_fill_manual(values = c("turquoise3", "magenta3", "skyblue4"),
labels = c("posterior_probability_relapse" = "Relapse",
"posterior_probability_recrudescence" = "Recrudescence",
"posterior_probability_reinfection" = "Reinfection")) +
scale_x_discrete(expand = c(0, 0)) +
scale_y_continuous(expand = c(0, 0)) +
labs(x = "Sample pair",
y = "Posterior marginal probability",
fill = "") +
theme_bw() +
facet_wrap(~ibd_range, scales = "free_x") +
theme(axis.text.x = element_text(angle = 90, hjust = 1, vjust = 0.5))
Plot IBD range vs probabilistic classification (joint probability, highest ranking)
classification_summary %>%
select(subject_id, sample_id, pair, episode_number, visit_date, days_since_enrolment, days_since_last_episode, starts_with("joint"), ibd_range, CI_lower, CI_upper, sig_est) %>%
ggplot(aes(x = reorder(pair, episode_number), y = joint_probability, group = episode_number, fill = joint_classification)) +
geom_bar(stat = "identity") +
scale_fill_manual(values = c("Relapse" = "turquoise3",
"Recrudescence" = "skyblue4",
"Reinfection" = "magenta3")) +
scale_x_discrete(expand = c(0, 0)) +
scale_y_continuous(expand = c(0, 0)) +
labs(x = "Sample pair",
y = "Posterior joint probability (highest ranking classification)",
fill = "") +
theme_bw() +
facet_wrap(~ibd_range, scales = "free_x") +
theme(axis.text.x = element_text(angle = 90, hjust = 1, vjust = 0.5))
Confusion matrices: PNH vs probabilistic classification
Classification by 1-PNH (IBS) range
confusion_data_ibsrange <- classification_summary %>%
select(subject_id, sample_id, episode_number, visit_date, days_since_enrolment, days_since_last_episode, starts_with("posterior_"), starts_with("joint"), prop_same_haps_1minusPNH, pnh_range, ibs_range,
# renaming for clarity
pnh_classification = classification) %>%
pivot_longer(cols = (starts_with("posterior")),
names_to = "marginal_classification",
values_to = "marginal_probability") %>%
mutate(marginal_classification = case_when(marginal_classification == "posterior_probability_recrudescence" ~ "Recrudescence",
marginal_classification == "posterior_probability_relapse" ~ "Relapse",
marginal_classification == "posterior_probability_reinfection" ~ "Reinfection"),
marginal_classification = factor(marginal_classification,
levels = c("Relapse", "Recrudescence", "Reinfection")),
marginal_probability_range = cut(marginal_probability,
breaks = c(0, 0.25, 0.5, 0.75, 1),
include.lowest = TRUE,
labels = c("0-0.25", "0.25-0.5", "0.5-0.75", "0.75-1")),
joint_probability_range = cut(joint_probability,
breaks = c(0, 0.25, 0.5, 0.75, 1),
include.lowest = TRUE,
labels = c("0-0.25", "0.25-0.5", "0.5-0.75", "0.75-1"))) %>%
select(subject_id, episode_number, days_since_enrolment, ibs_range, pnh_classification, marginal_classification, marginal_probability, marginal_probability_range, joint_classification, joint_probability, joint_probability_range) %>%
# now only keep the highest 'ranking' marginal classification
group_by(subject_id, episode_number) %>%
slice_max(order_by = marginal_probability, n = 1, with_ties = F) %>%
ungroup() %>%
group_by(ibs_range, marginal_classification, joint_classification) %>%
tally() %>%
dcast(ibs_range ~ marginal_classification + joint_classification, value.var = "n", fill = 0)confusion_data_ibsrange %>%
melt(id.vars = "ibs_range") %>%
ggplot(aes(ibs_range, variable, fill = value)) +
geom_tile() +
scale_fill_gradient(low = "white", high = "red") +
geom_text(aes(label = value)) +
theme_minimal() +
labs(x = "IBS range (1-PNH)",
y = "Marginal vs joint classification",
fill = "Count")
Classification comparison
# Create a confusion matrix for each pair
confusion_data_pnhstate <-
classification_summary %>%
select(subject_id, sample_id, episode_number, visit_date, days_since_enrolment, days_since_last_episode, starts_with("posterior_"), starts_with("joint"), prop_same_haps_1minusPNH, pnh_range, ibs_range,
# renaming for clarity
pnh_classification = classification) %>%
pivot_longer(cols = (starts_with("posterior")),
names_to = "marginal_classification",
values_to = "marginal_probability") %>%
mutate(marginal_classification = case_when(marginal_classification == "posterior_probability_recrudescence" ~ "Recrudescence",
marginal_classification == "posterior_probability_relapse" ~ "Relapse",
marginal_classification == "posterior_probability_reinfection" ~ "Reinfection"),
marginal_classification = factor(marginal_classification,
levels = c("Relapse", "Recrudescence", "Reinfection")),
marginal_probability_range = cut(marginal_probability,
breaks = c(0, 0.25, 0.5, 0.75, 1),
include.lowest = TRUE,
labels = c("0-0.25", "0.25-0.5", "0.5-0.75", "0.75-1")),
joint_probability_range = cut(joint_probability,
breaks = c(0, 0.25, 0.5, 0.75, 1),
include.lowest = TRUE,
labels = c("0-0.25", "0.25-0.5", "0.5-0.75", "0.75-1")),
# pnh_state = case_when(pnh_classification == "Heterologous" ~ "Reinfection",
# pnh_classification == "Homologous" ~ "Relapse or Recrudescence")
pnh_state = case_when(ibs_range == "0-0.25" | ibs_range == "0.25-0.5" ~ "Reinfection",
(ibs_range == "0.5-0.75" | ibs_range == "0.75-1") ~ "Relapse or Recrudescence")
) %>%
select(subject_id, episode_number, days_since_enrolment, ibs_range, pnh_classification, pnh_state, marginal_classification, marginal_probability, marginal_probability_range, joint_classification, joint_probability, joint_probability_range) %>%
# now only keep the highest 'ranking' marginal classification
group_by(subject_id, episode_number) %>%
slice_max(order_by = marginal_probability, n = 1, with_ties = F) %>%
ungroup() %>%
group_by(pnh_state, marginal_classification, joint_classification) %>%
tally() %>%
dcast(pnh_state ~ marginal_classification + joint_classification, value.var = "n", fill = 0)confusion_data_pnhstate %>%
melt(id.vars = "pnh_state") %>%
ggplot(aes(pnh_state, variable, fill = value)) +
geom_tile() +
scale_fill_gradient(low = "white", high = "red") +
geom_text(aes(label = value)) +
theme_minimal() +
labs(x = "PNH classification",
y = "Marginal vs joint classification",
fill = "Count")
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